Intelligent Computing and Optimization: Proceedings of the 6th International Conference on Intelligent Computing and Optimization 2023 (ICO2023), Volume 5 (Lecture Notes in Networks and Systems) 3031501578, 9783031501579

Citation preview

Lecture Notes in Networks and Systems 855

Pandian Vasant · Mohammad Shamsul Arefin · Vladimir Panchenko · J. Joshua Thomas · Elias Munapo · Gerhard-Wilhelm Weber · Roman Rodriguez-Aguilar   Editors

Intelligent Computing and Optimization Proceedings of the 6th International Conference on Intelligent Computing and Optimization 2023 (ICO2023), Volume 5

Lecture Notes in Networks and Systems

855

Series Editor Janusz Kacprzyk , Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland

Advisory Editors Fernando Gomide, Department of Computer Engineering and Automation—DCA, School of Electrical and Computer Engineering—FEEC, University of Campinas—UNICAMP, São Paulo, Brazil Okyay Kaynak, Department of Electrical and Electronic Engineering, Bogazici University, Istanbul, Türkiye Derong Liu, Department of Electrical and Computer Engineering, University of Illinois at Chicago, Chicago, USA Institute of Automation, Chinese Academy of Sciences, Beijing, China Witold Pedrycz, Department of Electrical and Computer Engineering, University of Alberta, Alberta, Canada Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Marios M. Polycarpou, Department of Electrical and Computer Engineering, KIOS Research Center for Intelligent Systems and Networks, University of Cyprus, Nicosia, Cyprus Imre J. Rudas, Óbuda University, Budapest, Hungary Jun Wang, Department of Computer Science, City University of Hong Kong, Kowloon, Hong Kong

The series “Lecture Notes in Networks and Systems” publishes the latest developments in Networks and Systems—quickly, informally and with high quality. Original research reported in proceedings and post-proceedings represents the core of LNNS. Volumes published in LNNS embrace all aspects and subfields of, as well as new challenges in, Networks and Systems. The series contains proceedings and edited volumes in systems and networks, spanning the areas of Cyber-Physical Systems, Autonomous Systems, Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power Systems, Robotics, Social Systems, Economic Systems and other. Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution and exposure which enable both a wide and rapid dissemination of research output. The series covers the theory, applications, and perspectives on the state of the art and future developments relevant to systems and networks, decision making, control, complex processes and related areas, as embedded in the fields of interdisciplinary and applied sciences, engineering, computer science, physics, economics, social, and life sciences, as well as the paradigms and methodologies behind them. Indexed by SCOPUS, INSPEC, WTI Frankfurt eG, zbMATH, SCImago. All books published in the series are submitted for consideration in Web of Science. For proposals from Asia please contact Aninda Bose ([email protected]).

Pandian Vasant · Mohammad Shamsul Arefin · Vladimir Panchenko · J. Joshua Thomas · Elias Munapo · Gerhard-Wilhelm Weber · Roman Rodriguez-Aguilar Editors

Intelligent Computing and Optimization Proceedings of the 6th International Conference on Intelligent Computing and Optimization 2023 (ICO2023), Volume 5

Editors Pandian Vasant Faculty of Electrical and Electronics Engineering, Modeling Evolutionary Algorithms Simulation and Artificial Intelligence Ton Duc Thang University Ho Chi Minh City, Vietnam Vladimir Panchenko Laboratory of Non-traditional Energy Systems, Department of Theoretical and Applied Mechanics, Federal Scientific Agroengineering Center VIM Russian University of Transport Moscow, Russia

Mohammad Shamsul Arefin Department of Computer Science Chittagong University of Engineering and Technology Chittagong, Bangladesh J. Joshua Thomas Department of Computer Science UOW Malaysia KDU Penang University College George Town, Malaysia Gerhard-Wilhelm Weber Faculty of Engineering Management Pozna´n University of Technology Pozna´n, Poland

Elias Munapo School of Economics and Decision Sciences North West University Mmabatho, South Africa Roman Rodriguez-Aguilar Facultad de Ciencias Económicas y Empresariales, School of Economic and Business Sciences Universidad Panamericana Mexico City, Mexico

ISSN 2367-3370 ISSN 2367-3389 (electronic) Lecture Notes in Networks and Systems ISBN 978-3-031-50157-9 ISBN 978-3-031-50158-6 (eBook) https://doi.org/10.1007/978-3-031-50158-6 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland Paper in this product is recyclable.

Preface

The sixth edition of the International Conference on Intelligent Computing and Optimization (ICO’2023) was held during April 27–28, 2023, at G Hua Hin Resort and Mall, Hua Hin, Thailand. The objective of the international conference is to bring the global research scholars, experts and scientists in the research areas of intelligent computing and optimization from all over the world to share their knowledge and experiences on the current research achievements in these fields. This conference provides a golden opportunity for global research community to interact and share their novel research results, findings and innovative discoveries among their colleagues and friends. The proceedings of ICO’2023 is published by SPRINGER (in the book series Lecture Notes in Networks and Systems) and indexed by SCOPUS. Almost 70 authors submitted their full papers for the 6th ICO’2023. They represent more than 30 countries, such as Australia, Bangladesh, Bhutan, Botswana, Brazil, Canada, China, Germany, Ghana, Hong Kong, India, Indonesia, Japan, Malaysia, Mauritius, Mexico, Nepal, the Philippines, Russia, Saudi Arabia, South Africa, Sri Lanka, Thailand, Turkey, Ukraine, UK, USA, Vietnam, Zimbabwe and others. This worldwide representation clearly demonstrates the growing interest of the global research community in our conference series. The organizing committee would like to sincerely thank all the authors and the reviewers for their wonderful contribution for this conference. The best and high-quality papers will be selected and reviewed by International Program Committee in order to publish the extended version of the paper in the international indexed journals by SCOPUS and ISI WoS. This conference could not have been organized without the strong support and help from LNNS SPRINGER NATURE, Easy Chair, IFORS and the Committee of ICO’2023. We would like to sincerely thank Prof. Roman Rodriguez-Aguiler (Universidad Panamericana, Mexico) and Prof. Mohammad Shamsul Arefin (Daffodil International University, Bangladesh), Prof. Elias Munapo (North West University, South Africa) and Prof. José Antonio Marmolejo Saucedo (National Autonomous University of Mexico, Mexico) for their great help and support for this conference. We also appreciate the wonderful guidance and support from Dr. Sinan Melih Nigdeli (Istanbul University—Cerrahpa¸sa, Turkey), Dr. Marife Rosales (Polytechnic University of the Philippines, Philippines), Prof. Rustem Popa (Dunarea de Jos University, Romania), Prof. Igor Litvinchev (Nuevo Leon State University, Mexico), Dr. Alexander Setiawan (Petra Christian University, Indonesia), Dr. Kreangkri Ratchagit (Maejo University, Thailand), Dr. Ravindra Boojhawon (University of Mauritius, Mauritius), Prof. Mohammed Moshiul Hoque (CUET, Bangladesh), Er. Aditya Singh (Lovely Professional University, India), Dr. Dmitry Budnikov (Federal Scientific Agroengineering Center VIM, Russia), Dr. Deepanjal Shrestha (Pokhara University, Nepal), Dr. Nguyen Tan Cam (University of Information Technology, Vietnam) and Dr. Thanh Dang Trung (Thu Dau Mot University, Vietnam). The ICO’2023 committee would like to sincerely thank all the authors, reviewers, keynote speakers (Prof. Roman Rodriguez-Aguiler,

vi

Preface

Prof. Kaushik Deb, Prof. Rolly Intan, Prof. Francis Miranda, Dr. Deepanjal Shrestha, Prof. Sunarin Chanta), plenary speakers (Prof. Celso C. Ribeiro, Prof. José Antonio Marmolejo, Dr. Tien Anh Tran), session chairs and participants for their outstanding contribution to the success of the 6th ICO’2023 in Hua Hin, Thailand. Finally, we would like to sincerely thank Prof. Dr. Janusz Kacprzyk, Dr. Thomas Ditzinger, Dr. Holger Schaepe and Ms. Varsha Prabakaran of LNNS SPRINGER NATURE for their great support, motivation and encouragement in making this event successful in the global stage. April 2023

Dr. Pandian Vasant (Chair) Prof. Dr. Gerhard-Wilhelm Weber Prof. Dr. Mohammad Shamsul Arefin Prof. Dr. Roman Rodriguez-Aguiler Dr. Vladimir Panchenko Prof. Dr. Elias Munapo Dr. J. Joshua Thomas

Contents

Business, Economics, Finance, Management, Social and Smart Technology Validating the Measurement Scale Items on Readiness to Adopt Human Resource Analytics in the Organizations of Nepal . . . . . . . . . . . . . . . . . . . . . . . . . . Shanti Devi Chhetri, Devesh Kumar, Deepesh Ranabhat, and Pradeep Sapkota Optimizing Fire Response Unit Location for Urban-Rural Area . . . . . . . . . . . . . . Sunarin Chanta and Ornurai Sangsawang Effects of Financial Literacy on Financial Inclusion: Evidence from Nepal’s Gandaki Province . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Deepesh Ranabhat, Narinder Verma, Pradeep Sapkota, and Shanti Devi Chhetri Business and Information Technology Strategy Impact on Organizational Performance: A Case Study of Nepal Telecom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sudip Poudel, Neesha Rajkarnikar, Deepanjal Shrestha, Deepmala Shrestha, and Seung Ryul Jeong A Systematic Literature Review on Factors Affecting Rural Tourism . . . . . . . . . . Pradeep Sapkota, Kamal Kant Vashisth, and Deepesh Ranabhat K-Modes with Binary Logistic Regression: An Application in Marketing Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jonathan Rebolledo and Roman Rodriguez-Aguilar Optimizing Headways Using Evolutionary Algorithms for Improved Light Rail Transit System Efficiency and Passenger Service . . . . . . . . . . . . . . . . . . . . . . . Oomesh Gukhool, Nooswaibah Binti Nooroodeen Soosor, Ravindra Boojhawon, and Mohammad Zaheer Doomah Improving the Ergonomics of the Master Data Management System Using Annotated Metagraph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. R. Nikolsky, A. A. Sukhobokov, and Goryachkin B.S. Patent Classification for Business Strategy with BERT . . . . . . . . . . . . . . . . . . . . . . Masaki Higashi, Yoshimasa Utsumi, and Kazuhide Nakata

3

14

22

33

45

56

65

76

84

viii

Contents

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria Decision Making Approach in Rapti River Watershed, India . . . . . . . . . . . . . . . . . Raashid Khan, Jawed Anwar, Saif said, Sarfarazali Ansari, Azazkhan Ibrahimkhan Pathan, and Lariyah Mohd Sidek

95

Optimizing Laser Drilling of Kenaf/HDPE Composites: A Novel CRITIC-MABAC Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 Sellamuthu Prabhukumar, Jasgurpeet Singh Chohan, and Kanak Kalita Education, Healthcare, Industry, and Advanced Engineering Determination of the Optimal Speed of Movement of the Conveyor Belt of the Prototype Weighing Belt Batcher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Denis Shilin, Dmitry Shestov, Alexey Vasiliev, and Valery Moskvin Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio City, Philippines from 2015 to 2018 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 Guinness G. Maza, Kendrick Jules G. Zante, Clarence Kyle L. Pagunsan, Angela Ronice A. Doctolero, Rostum Paolo B. Alanas, Criselda P. Libatique, and Rizavel C. Addawe The Economic Dimensions of the Non-communicable Diseases: A Panel Data Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 Sergio Arturo Domínguez-Miranda and Roman Rodriguez-Aguilar Re-strengthening of Real Sized RC Beams Subjected to Corrosion Using Glass Fiber Reinforced Polymer Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Sunil Garhwal, Shruti Sharma, Sandeep Kumar Sharma, Anil Garhwal, and Anirban Banik Optimization of the Lubricating and Cooling Fluid Composition . . . . . . . . . . . . . 153 I. Yu. Ignatkin, P. Kazantsev Sergey, D. M. Skorokhodov, N. V. Serov, T. Kildeev, A. V. Serov, and A. Anisimov Alexander Research of the Dosing Process with the Installation of Magnetic Stimulation of Seeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 V. Syrkin, S. Mashkov, P. Ishkin, S. Vasilev, and Yu. Daus Investigation of Hydrodynamic Behaviour in Rectangular Sheet Shaped Membrane Using Computational Fluid Dynamics (CFD) . . . . . . . . . . . . . . . . . . . . 170 Anirban Banik, Sushant Kumar Biswal, Tarun Kanti Bandyopadhyay, Vladimir Panchenko, Sunil Garhwal, and Anil Garhwal

Contents

ix

A Review on the Impacts of Social Media on the Mental Health . . . . . . . . . . . . . . 181 Md. Abu Bakar Siddiq Tapu, Rashik Shahriar Akash, Hafiz Al Fahim, Tanin Mohammad Jarin, Touhid Bhuiyan, Ahmed Wasif Reza, and Mohammad Shamsul Arefin Factor Influencing Online Purchase Intention Among University Students in Nepal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Deepesh Ranabhat, Sujita Adhikari, and Narinder Verma Effect of Thin Polymer Interlayers in the Spindle-Bearing Joint on the Stiffness and Durability of Spindle Bearing Assemblies of Mills . . . . . . . . 207 A. S. Kononenko, T. A. Kildeev, Ignatkin I. Yu, and N. A. Sergeeva The Use of a Nutrient Solution Containing Chelated Forms of Various Trace Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 K. Pishchaeva, S. Muradyan, E. Nikulina, S. Buleeva, and A. Saproshina Designing an Inventory Control System in Food and Beverage Industry . . . . . . . 223 Tiovitus Flomando Tunga and Tanti Octavia Evaluating Research Impact: A Comprehensive Overview of Metrics and Online Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 Seema Ukidve, Ramsagar Yadav, Mukhdeep Singh Manshahia, and Jasleen Randhawa Hazard Identification, Risk Assessment and Control (HIRAC) at the Wood Processing Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 Herry Christian Palit and Alexander Location Selection of Rail Transportation Hub Using TOPSIS Method . . . . . . . . 253 Kanokporn Sripathomswat, Nattawat Tipchareon, Worapat Aruntippaitoon, Itiphong Trirattanasarana, and Sunarin Chanta Developing a Transaction System in Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . 262 Afsana Nur Meem, Lamya Ishrat Nodi, Efte Kharul Islam, Minhazul Amin Tomal, Ahmed Wasif Reza, and Mohammad Shamsul Arefin Using the Phi-Function Technique for the Optimized Virtual Localization Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 Sergiy Plankovskyy, Yevgen Tsegelnyk, Tetyana Romanova, Oleksandr Pankratov, Igor Litvinchev, and Volodymyr Kombarov

x

Contents

COVID-19 Detection from Chest X-Ray Images Using CNN Models and Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 Nafisha Binte Moin, Shamima Sultana, Abdullah Al Munem, Omar Tawhid Imam, Ahmed Wasif Reza, and Mohammad Shamsul Arefin A Note on Solving the Transportation Model by the Hungarian Method of Assignment: Unification of the Transportation and Assignment Models . . . . . 301 Santosh Kumar, Trust Tawanda, Elias Munapo, and Philimon Nyamugure Automatic Crack Detection Approach for the Offshore Flare System Inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314 Teepakorn Tosawadi, Pakcheera Choppradit, Satida Sookpong, Sasin Phimsiri, Vasin Suttichaya, Chaitat Utintu, and Ek Thamwiwatthana The Recent Trend of Artificial Neural Network in the Field of Civil Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324 Aditya Singh Analyzing Price Forecasting of Grocery Products in Bangladesh: A Comparison of Time Series Modeling Approaches . . . . . . . . . . . . . . . . . . . . . . . 334 Md Mahmudul Hoque, Ikbal Ahmed, Nayan Banik, and Mohammed Moshiul Hoque Big Data: Identification of Critical Success Factors . . . . . . . . . . . . . . . . . . . . . . . . . 342 Leo Willyanto Santoso Effect of Parameter Value of a Hybrid Algorithm for Optimization of Truss Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 Melda Yücel, Sinan Melih Nigdeli, and Gebrail Bekda¸s Brain MRI Classification for Alzheimer’s Disease Based on Convolutional Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 Md. Saiful, Arpita Saha, Faria Tabassum Mim, Nafisa Tasnim, Ahmed Wasif Reza, and Mohammad Shamsul Arefin Drivers and Barriers for Going Paperless in Tertiary Educational Institute . . . . . . 368 Rafid Mahmud Haque, Lamyea Tasneem Maha, Oshin Nusrat Rahman, Noor Fabi Shah Safa, Rashedul Amin Tuhin, Ahmed Wasif Reza, and Mohammad Shamsul Arein Impact of Lifestyle on Career: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Md. Jabed Hosen, Md. Injamul Haque, Saiful Islam, Mohammed Nadir Bin Ali, Touhid Bhuiyan, Ahmed Wasif Reza, and Mohammad Shamsul Arefin

Contents

xi

Deep Learning Approach for COVID-19 Detection: A Diagnostic Tool Based on VGG16 and VGG19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394 Fardin Rahman Akash, Ajmiri Afrin Priniya, Jahani Shabnam Chadni, Jobaida Ahmed Shuha, Ismot Ara Emu, Ahmed Wasif Reza, and Mohammad Shamsul Arefin A Survey of Modeling the Healthcare Inventory for Emerging Infectious Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404 Tatitayakorn Limsakul and Sompoap Taladgaew MEREC-MABAC Based-Parametric Optimization of Chemical Vapour Deposition Process for Diamond-Like Carbon Coatings . . . . . . . . . . . . . . . . . . . . . 414 Sellamuthu Prabhukumar, Jasgurpeet Singh Chohan, and Kanak Kalita Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423

About the Editors

Pandian Vasant is Research Associate at Modeling Evolutionary Algorithms Simulation and Artificial Intelligence, Faculty of Electrical and Electronics Engineering, Ton Duc Thang University, Ho Chi Minh City, Vietnam, and Editor in Chief of International Journal of Energy Optimization and Engineering (IJEOE). He holds Ph.D. in Computational Intelligence (UNEM, Costa Rica), M.Sc. (University Malaysia Sabah, Malaysia, Engineering Mathematics) and B.Sc. (Hons, Second Class Upper) in Mathematics (University of Malaya, Malaysia). His research interests include soft computing, hybrid optimization, innovative computing and applications. He has co-authored research articles in journals, conference proceedings, presentations, special issues Guest Editor, chapters (312 publications indexed in Research-Gate) and General Chair of EAI International Conference on Computer Science and Engineering in Penang, Malaysia (2016) and Bangkok, Thailand (2018). In the years 2009 and 2015, he was awarded top reviewer and outstanding reviewer for the journal Applied Soft Computing (Elsevier). He has 35 years of working experience at the universities. Currently, Dr. Pandian Vasant is General Chair of the International Conference on Intelligent Computing and Optimization (https://www.icico.info/) and Research Associate at Modeling Evolutionary Algorithms Simulation and Artificial Intelligence, Faculty of Electrical and Electronics Engineering, Ton Duc Thang University, HCMC, Vietnam. Professor Dr. Mohammad Shamsul Arefin is in lien from Chittagong University of Engineering and Technology (CUET), Bangladesh and currently affiliated with the Department of Computer Science and Engineering (CSE), Daffodil International University (DIU), Dhaka, Bangladesh. Earlier he was the head of CSE Department, CUET. Prof. Arefin received his Doctor of Engineering Degree in Information Engineering from Hiroshima University, Japan with support of the scholarship of MEXT, Japan. As a part of his doctoral research, Dr. Arefin was with IBM Yamato Software Laboratory, Japan. His research includes data privacy and mining, big data management,

xiv

About the Editors

IoT, Cloud Computing, Natural Language processing, Image Information Processing, Social Networks Analysis and Recommendation Systems and IT for agriculture, education and environment. Prof. Arefin is the Editor in Chief of Computer Science and Engineering Research Journal (ISSN: 1990-4010) and was the Associate Editor of BCS Journal of Computer and Information Technology (ISSN: 2664-4592) and a reviewer as well as TPC member of many international journals and conferences. Dr. Arefin has more than 120 referred publications in international journals, book series and conference proceedings. He delivered more than 30 keynote speeches/invited talks. He also received a good number of research grants/funds from home and abroad. Dr. Arefin is a senior member of IEEE, Member of ACM, Fellow of IEB and BCS. Prof. Arefin involves/earlier involved in many professional activities such as Chairman of Bangladesh Computer Society (BCS) Chittagong Branch; Vice-President (Academic) of BCS National Committee; Executive Committee Member of IEB Computer Engineering Division; Advisor, Bangladesh Robotic Foundation. He was also a member of pre-feasibility study team of CUET IT Business Incubator, first campus based IT Business Incubator in Bangladesh. Prof. Arefin is an Principle Editor of the Lecture Notes on Data Engineering and Communications Technologies book series (LNDECT, Volume 95) published by Springer and an editor of the books on Applied Informatics for Industry 4.0, Applied Intelligence for Industry 4.0 and Computer Vision and Image Analysis for Industry 4.0 to be published Tailor and Francis. Prof. Arefin is the Vice-Chair (Technical) of IEEE CS BDC for the year 2022. He was the Vice-Chair (Activity) of IEEE CS BDC for the year 2021 and the Conference Co-Coordinator of IEEE CS BDC for two consecutive years, 2018 and 2019. He is acting as a TPC Chair of MIET 2022 and the TPC Chair of IEEE Summer Symposium 2022. He was the Organizing Chair of International Conference on Big Data, IoT and Machine Learning (BIM 2021) and National Workshop on Big Data and Machine Learning (BDML 2020). He served as the TPC Chair, International Conference on Electrical, Computer and Communication Engineering (ECCE 2017); Organizing Co-chair, ECCE 2019, Technical Co-chair, IEEE CS BDC Winter Symposium 2020 and Technical Secretary, IEEE CS BDC Winter Symposium 2021. Dr. Arefin helped different international conferences

About the Editors

xv

in the form of track chair, TPC member, reviewer and/or secession chair etc. He is a reviewer of many reputed journals including IEEE Access, Computing Informatics, ICT Express, Cognitive Computation etc. Dr. Arefin visited Japan, Indonesia, Malaysia, Bhutan, Singapore, South Korea, Egypt, India, Saudi Arabia and China for different professional and social activities. Vladimir Panchenko is an Associate Professor of the “Department of Theoretical and Applied Mechanics” of the “Russian University of Transport”, Senior Researcher of the “Laboratory of Non-traditional Energy Systems” of the “Federal Scientific Agroengineering Center VIM” and the Teacher of additional education. Graduated from the “Bauman Moscow State Technical University” in 2009 with the qualification of an engineer. Ph.D. thesis of the specialty “Power plants based on renewable energy” was defended in 2013. From 2014 to 2016 Chairman of the Council of Young Scientists and the Member of the Academic Council of the All-Russian Institute for Electrification of Agriculture, Member of the Council of Young Scientists of the Russian University of Transport, Member of the International Solar Energy Society, Individual supporter of Greenpeace and the World Wildlife Fund, Member of the Russian Geographical Society, Member of the Youth section of the Council “Science and Innovations of the Caspian Sea”, Member of the Committee on the use of renewable energy sources of the Russian Union of Scientific and Engineering Public Associations. Diplomas of the winner of the competition of works of young scientists of the AllRussian Scientific Youth School with international participation “Renewable Energy Sources”, Moscow State University M.V. Lomonosov in 2012, 2014, 2018 and 2020, Diploma with a bronze medal of the 15th Russian agroindustrial exhibition “Golden Autumn—2013”, Diploma with a gold medal of the 18th Russian agro-industrial exhibition “Golden Autumn—2016”, Diploma with a silver medal of the XIX Moscow International Salon of Inventions and Innovative technologies “Archimedes—2016”, Diploma for the winning the schoolchildren who have achieved high results in significant events of the Department of Education and Science of the City of Moscow (2020–2021, School No. 2045). Scientific adviser of schoolchildren-winners and prize-winners of the Project and Research Competition “Engineers of the Future” at NUST MISiS 2021 and

xvi

About the Editors

RTU MIREA 2022. Invited expert of the projects of the final stages of the “Engineers of the Future” (2021, 2022) and the projects of the “Transport of the Future” (2022, Russian University of Transport). Grant “Young teacher of MIIT” after competitive selection in accordance with the Regulations on grants for young teachers of MIIT (2016– 2019). Scholarship of the President of the Russian Federation for 2018–2020 for young scientists and graduate students carrying out promising research and development in priority areas of modernization of the Russian economy. Grant of the Russian Science Foundation 2021 “Conducting fundamental scientific research and exploratory scientific research by international research teams”. Reviewer of articles, chapters and books IGI, Elsevier, Institute of Physics Publishing, International Journal of Energy Optimization and Engineering, Advances in Intelligent Systems and Computing, Journal of the Operations Research Society of China, Applied Sciences, Energies, Sustainability, AgriEngineering, Ain Shams Engineering Journal, Concurrency and Computation: Practice and Experience. Presenter of the sections of the Innovations in Agriculture conference, keynote speaker of the ICO 2019 conference session, keyspeaker of the special session of the ICO 2020 conference. Assistant Editor since 2019 of the “International Journal of Energy Optimization and Engineering”, Guest Editor since 2019 of the Special Issues of the journal MDPI (Switzerland) “Applied Sciences”, Editor of the book of the “IGI GLOBAL” (USA), as well as book of the “Nova Science Publisher” (USA). Participated in more than 100 exhibitions and conferences of various levels. Published more than 250 scientific papers, including 14 patents, 1 international patent, 6 educational publications, 4 monographs. J. Joshua Thomas is an Associate Professor at UOW Malaysia KDU Penang University College, Malaysia since 2008. He obtained his Ph.D. (Intelligent Systems Techniques) in 2015 from University Sains Malaysia, Penang, and Master’s degree in 1999 from Madurai Kamaraj University, India. From July to September 2005, he worked as a research assistant at the Artificial Intelligence Lab in University Sains Malaysia. From March 2008 to March 2010, he worked as a research associate at the same University. Currently, he is working with Machine Learning, Big Data, Data Analytics, Deep Learning, specially targeting on

About the Editors

xvii

Convolutional Neural Networks (CNN) and Bi-directional Recurrent Neural Networks (RNN) for image tagging with embedded natural language processing, End to end steering learning systems and GAN. His work involves experimental research with software prototypes and mathematical modelling and design He is an editorial board member for the Journal of Energy Optimization and Engineering (IJEOE), and invited guest editor for Journal of Visual Languages Communication (JVLC-Elsevier). Recently with Computer Methods and Programs in Biomedicine (Elsevier). He has published more than 40 papers in leading international conference proceedings and peer reviewed journals. Elias Munapo has a Ph.D. obtained in 2010 from the National University of Science and Technology (Zimbabwe) and is a Professor of Operations Research at the North West University, Mafikeng Campus in South Africa. He is a Guest Editor of the Applied Sciences Journal and has co-published two books. The first book is titled Some Innovations in OR Methodology: Linear Optimization and was published by Lambert Academic publishers in 2018. The second book is titled Linear Integer Programming: Theory, Applications, and Recent Developments and was published by De Gruyter publishers in 2021. Professor Munapo has co-edited a number of books, is currently a reviewer of a number of journals, and has published over 100 journal articles and book chapters. In addition, Prof. Munapo is a recipient of the North West University Institutional Research Excellence award and is a member of the Operations Research Society of South Africa (ORSSA), EURO, and IFORS. He has presented at both local and international conferences and has supervised more than 10 doctoral students to completion. His research interests are in the broad area of operations research.

xviii

About the Editors

Gerhard-Wilhelm Weber is a Professor at Poznan University of Technology, Poznan, Poland, at Faculty of Engineering Management. His research is on mathematics, statistics, operational research, data science, machine learning, finance, economics, optimization, optimal control, management, neuro-, bio- and earth-sciences, medicine, logistics, development, cosmology and generalized spacetime research. He is involved in the organization of scientific life internationally. He received Diploma and Doctorate in Mathematics, and Economics/Business Administration, at RWTH Aachen, and Habilitation at TU Darmstadt (Germany). He replaced Professorships at University of Cologne, and TU Chemnitz, Germany. At Institute of Applied Mathematics, Middle East Technical University, Ankara, Turkey, he was a Professor in Financial Mathematics and Scientific Computing, and Assistant to the Director, and has been a member of five further graduate schools, institutes and departments of METU. G.-W. Weber has affiliations at Universities of Siegen (Germany), Federation University (Ballarat, Australia), University of Aveiro (Portugal), University of North Sumatra (Medan, Indonesia), Malaysia University of Technology, Chinese University of Hong Kong, KTO Karatay University (Konya, Turkey), Vidyasagar University (Midnapore, India), Mazandaran University of Science and Technology (Babol, Iran), Istinye University (Istanbul, Turkey), Georgian International Academy of Sciences, at EURO (Association of European OR Societies) where he is “Advisor to EURO Conferences” and IFORS (International Federation of OR Societies), where he is member in many national OR societies, honorary chair of some EURO working groups, subeditor of IFORS Newsletter, member of IFORS Developing Countries Committee, of Pacific Optimization Research Activity Group, etc. G.-W. Weber has supervised many M.Sc. and Ph.D. students, authored and edited numerous books and articles, and given many presentations from a diversity of areas, in theory, methods and practice. He has been a member of many international editorial, special issue and award boards; he participated at numerous research projects; he received various recognitions by students, universities, conferences and scientific organizations. G.-W. Weber is an IFORS Fellow.

About the Editors

xix

Roman Rodriguez-Aguilar is a professor in the School of Economic and Business Sciences of the “Universidad Panamericana” in Mexico. His research is on large-scale mathematical optimization, evolutionary computation, data science, statistical modeling, health economics, energy, competition, and market regulation. He is particularly interested in topics related to artificial intelligence, digital transformation, and Industry 4.0. He received his Ph.D. at the School of Economics at the National Polytechnic Institute, Mexico. He also has a master’s degree in Engineering from the School of Engineering at the National University of Mexico (UNAM), a master’s degree in Administration and Public Policy in the School of Government and Public Policy at Monterrey Institute of Technology and Higher Education, a postgraduate in applied statistics at the Research Institute in Applied Mathematics and Systems of the UNAM and his degree in Economics at the UNAM. Prior to joining Panamericana University, he has worked as a specialist in economics, statistics, simulation, finance, and optimization, occupying different management positions in various public entities such as the Ministry of Energy, Ministry of Finance, and Ministry of Health. At present, he has the secondhighest country-wide distinction granted by the Mexican National System of Research Scientists for scientific merit (SNI Fellow, Level 2). He has co-authored research articles in science citation index journals, conference proceedings, presentations, and book chapters.

Business, Economics, Finance, Management, Social and Smart Technology

Validating the Measurement Scale Items on Readiness to Adopt Human Resource Analytics in the Organizations of Nepal Shanti Devi Chhetri1(B) , Devesh Kumar1 , Deepesh Ranabhat2 and Pradeep Sapkota2

,

1 Faculty of Management Sciences, Shoolini University, Bajhol, Himachal Pradesh 173229,

India [email protected] 2 Faculty of Management Studies, Pokhara University, Pokhara, Nepal

Abstract. The key purpose of this study is to validate the measurement scale items on readiness to adopt Human Resource Analytics via pilot survey. Data were gathered using a questionnaire following the distribution of (30) questionnaires among the research sample. The data were examined with the use of the SPSS and SmartPLS programs. The total of 78 items were established out of 104 items for the testing reliability and validity of all the constructs. The study found that composite reliability, Cronbach’s Alpha, AVE, the heterotrait-monotrait (HTMT) ratio, Cross-loadings, and the Fornell and Larcker criterion, fulfill the required threshold for reliability and validity. The study concluded that further study can be done to validate the scale in the developing nations like Nepal. Keywords: Human resource analytics · Composite reliability · Discriminant validity · SmartPLS · Nepal

1 Introduction The most common organizational skill gap is still in HR analytics. The ideal environment for using HR data to make business and HR choices is being created by rapidly advancing technology capacity and increased accessibility to “people data” [1]. Real-time data analysis enables businesses to observe the past and predict the future. This is the benefit of streaming analytics, which can describe what happened, diagnose why it happened, predict what might happen in the future, and, ultimately, decide how to change the course of events (prescriptive). The corporate world is experiencing a widespread emergence of data analysis. Data is used widely in the sales, customer service, and financial sectors, and businesses are now looking to analyze data more in the human resources area as well [2]. Although analytics have long been used to inform choices about human capital management (HCM), far too many businesses still rely on these programs to present straightforward correlations and descriptive statistics. It still needs a lot of work before advanced analytics is completely employed for HCM decisions, although it has © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 3–13, 2024. https://doi.org/10.1007/978-3-031-50158-6_1

4

S. D. Chhetri et al.

been accepted by other corporate areas like finance and marketing [3]. Over the past ten years, data analytics has grown and changed by a variety of themes and discourses. The development and use of numerous academic models helped HR professionals adopt HR Analytics. In most companies, HR analytics are largely unheard of, therefore managers frequently rely on their gut instincts when assessing employee data. By explaining how HR practices and procedures affect an organization’s performance as a whole, HR analytics bring value to the company. Utilizing statistical tools and methods that assist in establishing a connection with HR practices and HR indicators constitutes HR analytics [4]. With the effective use of analytics in the human resources department, there may be a lot more seamless flow of communication between employers and employees, and the efficiency of employees can be measured. The main focus of this paper is to validate the measurement scale items on readiness to adopt HR analytics in the organizations of Nepal.

2 Literature Review 2.1 Importance of HR Analytics Analytics is the term used to describe the use of quantitative methodologies, decisionmaking, and computer science to the organization, analysis, and explication of the growing volume of data produced by contemporary civilization. The inclusion of the word “HR” denotes that these analyses are focused on internal company personnel [5]. However, [6] revealed that the majority of the surveyed organizations are at the basic level when it comes to human resource analytics, which means they use analytical tools that can measure HR based on descriptive factors, like the number of specialists who have received training and those who need it instead of figuring out a return on investment. Again, [7] explained that analytics data that may monitor employee attendance, workplace accidents, injuries, turnover costs, recruiting decisions, and engagement are used to forecast the future HR elements in the organization. It offers the organization a sustainable economic benefit. Also, [8] suggested that the productivity and profitability of an organization are likely to increase if it combines efficient HR management procedures with efficient HR technology. Akhmetova and Nevskaya [9] concluded that an organization’s operational effectiveness is improved via HR analytics. The effectiveness of data in decision-making, rather than the volume of data that is gathered and processed, determines the success of HR analytics. The HR division should not be the only one to consider HR analytics to be relevant. Analytics boosts a business’s overall performance. van den Heuvel and Bondarouk [10] discovered that by 2025, HR analytics will be a well-established field, have a measurable impact on business results, and play a key role in managerial and strategic decision-making. The integration will also be key to the development of HR analytics, linking data and IT infrastructure across disciplines and even corporate boundaries. The HR analytics position may also be a part of a central analytics role that crosses fields like marketing, finance, and HRM. Wandhe [11] stated that not only can HR analytics help firms make the best HR decisions possible with reliable data, but they also encourage them to keep appropriate quality data on hand to support their HR investment decisions. Although HR Analytics is a new field with considerable interest, the creation of an integrated, strategic framework for its implementation and

Validating the Measurement Scale Items on Readiness

5

operation will lessen the uncertainty that early adopters currently experience. Bakre [12] suggested that to give the best solutions for managing human capital, HR professionals must adapt to the changing circumstances and embrace an integrated management model. By determining the appropriate methods for gauging its influence on the success drivers, a business needs to comprehend and be able to measure the entire strategic impact of HR. 2.2 Factors Affecting Adoption of HR Analytics The researcher [13] used the UTAUT (the unified theory of technology adoption and use of technology) approach to evaluate construction organizations’ intentions to adopt Big Data Analytics (BDA). The study found that an organization’s decision to implement BDA is influenced by factors related to conducive conditions, performance expectations, and social influence. Likewise, [14] based on the Technology-Organization-Environment framework complexity, compatibility, governmental backing, size of the organization, level of competition, and environmental uncertainty were discovered to be important factors in determining the adoption of BDAs. Similarly, [15] investigated the factors affecting the intention to adopt BDA and found that competitive pressure, organizational data environment, relative advantage, complexity, compatibility, support from top-level management, technology readiness, and data availability in the organization significantly impact adoption. Also, [16] suggested that expectations of performance, effort, social influence, and facilitating circumstances influence the behavioral intention to adopt HR analytics. Lastly, [17] revealed that technological considerations, data infrastructure, and data quality control are the most crucial.

3 Research Methodology This paper focus on pilot study survey and adopts quantitative approach to analyze reliability and validity of the measurement items for the adoption of HR analytics in the organization of Nepal. The targeted sample for final study is 300 HR managers from different types of organization. Therefore, researcher distributed self-administered questionnaire to 30 HR managers working in different sectors through google forms and physical forms for pilot survey. Since, 10% of the total targeted sample is acceptable for the pilot study [18, 19]. The questionnaire is divided into three sections, the first section comprises personal details of the respondents, the second section is related to readiness and the third section includes 104 items on a 5-point Likert scale to analyze the constructs used in this study. Table 1 shows the details of the constructs, items and indicators used for the study. Data were analyzed using SPSS and Smart-PLS4. The personal details of respondents and their readiness to adopt HR analytics were analyzed in SPSS. Whereas the validity and reliability of the items were analyzed in Smart-PLS. The reliability and validity of the scale’s items were examined using Cronbach’s alpha coefficient, composite reliability, convergent and discriminant validity, and cross-loadings.

6

S. D. Chhetri et al. Table 1. List of constructs, items and indicators

Constructs

No. of items

Name of items

Indicators

Information technology infrastructure

8

ITI1, ITI2, ITI3, ITI4, ITI5, ITI6, ITI7, ITI8

Internet connection, technical infrastructure, advanced software

Relative advantage

6

RA1, RA2, RA3, RA4, RA5, RA6

Cost, quality of work, time, effort

Complexity

7

C1, C2, C3, C4, C5, C6, Easy to use, data C7 maintenance, implementation, analysis of data

Compatibility

7

CO1, CO2, CO3, CO4, CO5, CO6, CO7

Suitable, technical changes, match with present data, fits with organization

Top management support

6

TMS1, TMS2, TMS3, TMS4, TMS5, TMS6

Understand, support, involvement, encouragement

Learning organization

8

LO1, LO2, LO3, LO4, LO5, LO6, LO7, LO8

Advancements, acquisition of new skills, innovation

Financial readiness

6

FR1, FR2, FR3, FR4, FR5, FR6

Fund, budget allocation, loan

External pressure

6

EP1, EP2, EP3, EP4, EP5, EP6

Competition, response, competitive advantage

Employee orientation

5

EO1, EO2, EO3, EO4, EO5

Opinion, requirement, concern

Trading partner pressure

6

TP1, TP2, TP3, TP4, TP5, TP6

Recommendation, demand, request

Awareness

7

AW1, AW2, AW3, Information, data AW4, AW5, AW6, AW7 identification, advantage

Information technology expertise/analytical skills

8

ITA1, ITA2, ITA3, ITA4, ITA5, ITA6, ITA7, ITA8

Innovation

9

IN1, IN2, IN3, IN4, Learning, easy IN5, IN6, IN7, IN8, IN9 adoption, advancements

Technical knowledge, abilities, IT literate, proficiency

(continued)

Validating the Measurement Scale Items on Readiness

7

Table 1. (continued) Constructs

No. of items

Name of items

Indicators

Readiness

7

RS1, RS2, RS3, RS4, RS5, RS6, RS7

Ready to use, interested, learning, understanding

Level of adoption

8

HRA1, HRA2, HRA3, HRA4, HRA5, HRA6, HRA7, HRA8

Planning, initial phase, adopted, interested

4 Data Analysis and Results 4.1 Personal Profile and Readiness to Adopt HR Analytics The personal profile of respondents comprised of gender, type of organization they are employed in, qualification, and total experience as a human resource professional. Whereas, readiness to adopt HR analytics include no. of employees in the organization, the organization having a separate HR department, technology savvy human resource, collection of employee-related data, analysis of employee recorded data, separate team for HR analytics, benefits experience by using HR analytics and vision and mission for HR analytics. SPSS is used for analyzing the frequency distribution of respondent profiles and the status of the organization. In this paper, the majority of the respondents (66.7%) are male and only 33.3% of the respondents are females. However, the majority of the respondents (80%) are from service sectors which include banks, hospitals, cooperatives, and automobiles and only 20% of the respondents are from manufacturing sectors. In the case of education qualification, most of the respondents are having masters and above (83.3%), respondents hold a bachelor’s degree 13.3%, and only 3.3% hold a diploma certificate. Similarly, in terms of experience as an HR professional, most of the respondents (63.3%) fall under the category of 5 years and above. Likewise, the majority of the organization (73.3%) have employees more than 101. In terms of the organization having separate HR departments, most of them (76.7%) have separate departments. Also, most of the organizations (53.3%) are having 50% of the employees as technology savvy. Furthermore, the majority of the organizations (96.7%) collect employee-related data and most of them (90%) analyze employee-recorded data. Additionally, 76.7% of the organization have their team for the analysis of data. Moreover, most organizations (53.3%) can manage employee performance and productivity through HR analytics. Lastly, most of the organization (50%) have not yet planned for the vision and mission for HR analytics. 4.2 Reliability and Validity The validity and reliability is examined in the current study utilizing the partial least squares (PLS) method. The current study focused on measurement model of SmartPLS to validate the items in the measurement scale.

8

S. D. Chhetri et al.

Measurement Model The measurement model evaluates the connection between a construct and its observed elements. Indicator loading, construct reliability, convergent validity, and discriminant validity are examined during the assessment of a measurement model. To assess measurement model, study used 104 items in an initial phase (shown in Table 1). After removing 26 items (AW3, EO1, EO2, EO5, EP1, EP2, EP3, EP4, HRA3, HRA4, HRA5, IN6, IN7, IN8, IN9, ITI1, ITI2, ITI5, LO1, LO2, RS4, RS7, TP2, TP3, TP5, TP6), 78 total items were established for the testing reliability and validity of all the items. In the current study, information technology (IT) infrastructure, relative advantage, complexity, compatibility, support from top management, learning organization, financial readiness, external pressure, employees orientation, trading partner pressure, awareness, information technology expertise/analytical skills, innovation, readiness are considered as latent variables of HR analytics. Validating the Constructs Cronbach Alpha and Composite Reliability (CR) were used to examine the reliability of the study’s constructs. According to [20] both of these measures should have a value above 0.70 for the constructs to be reliable. The results in Table 2 show that both measures meet the requirements for acceptance, establishing reliability. Table 2. Composite reliability and convergent validity with acceptable values (Cronbach’s alpha, composite reliability ≥ 0.70 and AVE > 0.5) Constructs

Cronbach’s alpha

Composite reliability

(AVE)

Awareness

0.951

0.961

0.804

Compatibility

0.950

0.955

0.755

Complexity

0.912

0.927

0.647

Employee orientation

0.735

0.876

0.781

External pressure

0.793

0.884

0.794

Financial readiness

0.935

0.948

0.753

HRA adoption

0.874

0.909

0.666

IT infrastructure

0.962

0.970

0.865

IT expertise

0.957

0.963

0.765

Innovation

0.875

0.906

0.661

Learning organization

0.920

0.937

0.715

Readiness

0.897

0.923

0.706

Relative advantage

0.953

0.960

0.800

Top management support

0.956

0.964

0.818

Trading partner

0.909

0.957

0.917

Validating the Measurement Scale Items on Readiness

9

Construct Validity The degree to which a construct used in the study accurately measures what it is intended to assess is known as construct validity. Convergent validity and discriminant validity are two types of validity that are evaluated using construct validity. Convergent Validity Convergent validity is the measurement of the degree of agreement between several indicators of the same construct. The indicator’s cross-loadings, composite reliability (CR), and average variance extracted (AVE) all need to be analyzed to prove convergent validity. The value is between 0 and 1. To be sufficient for convergent validity, the AVE value should be more than 0.50 [21]. In the current study, all the measured constructs’ AVE value is above 0.50. Thus, convergent validity is determined. Discriminant Validity Discriminant validity describes how much an empirically determined construct differs from one another. It also evaluates the degree to which the overlapping constructions diverge from one another. Cross-loadings of indicators, the Fornell and Larcker criterion, and the heterotrait-monotrait (HTMT) ratio of correlation are used to evaluate the discriminant validity [20]. Cross Loadings When compared to other constructs in the study, an item should have larger loadings on its parent construct, according to cross-loadings [20]. The finding reveals that the cross-loadings of the items in their parent construct are higher than the loadings on other constructs. As a result, discriminant validity is determined. Fornell and Larcker In this approach, the correlation of latent constructs is contrasted with the average extracted square root of the variance (AVE). A latent construct should be able to account for variation in its indicator than for variance in other latent constructs. Accordingly, correlations with other latent constructs must be lower than the square root of each construct’s AVE [21]. According to Table 3, the square root of the AVE (top value in each column) was higher than its correlation with other components. The findings of the study provide strong support for discriminant validity. Heterotrait-Monotrait Ratio (HTMT) Ratio Correlations with an acceptable “Heterotrait-Monotrait ratio (HTMT)” value of less than 0.85 should be used to assess the discriminant validity [22]. The results in Table 4 demonstrate that all values are acceptable, hence the discriminant validity is established.

5 Conclusion The digital transformation of HR is a trend these days, emerging themes include Human Resource Analytics (HRA), artificial intelligence (AI), and cloud-based HR technology. However, gaps between research and practice are once again felt strongly. Developing

0.294

0.253

0.069

0.541

0.499

0.476

− 0.006

Employee orientation

External pressure

Financial readiness

HRA adoption

IT infrastructure

0.540

0.151

0.324

0.184

0.381

0.287

0.523

0.167

0.136

0.526

0.261

0.417

0.190

0.291

0.380

0.154

0.245

− 0.082

0.520

0.447

0.427

0.884

0.052

0.368

0.146

0.418

0.287

0.804

4

0.059

3

0.583

0.569

0.109

0.487

0.487

0.272

0.546

0.165

0.457

0.616

0.891

5

0.515

0.555

0.107

0.481

0.546

0.353

0.492

0.125

0.683

0.867

6

0.641

0.271

0.338

0.481

0.381

0.559

0.413

0.245

0.816

7

0.343

− 0.094

− 0.025

0.624

0.529

0.411

0.875

0.320

0.083

0.332

0.108

0.353

0.161

0.930

9

0.257

8

0.408

0.386

0.162

0.674

0.662

0.813

10

0.361

0.758

0.171

0.660

0.846

11

0.840

0.413

0.547

− 0.054

12

0.905

0.210

0.084

14

0.326

0.895

13

The bold values are square root of AVE which is higher than its correlation with other components which shows the establishment of discriminant validity

Trading partner

Relative advantage

0.259

0.676

− 0.105

Readiness

Top management support

0.385

0.517

0.462

Learning organization

0.397

0.472

Innovation

0.154

0.732

IT expertise

0.229

0.293

0.561

0.467

0.081

0.073

Complexity

0.869

2

Compatibility

0.897

1

Awareness

Table 3. Discriminant validity-Fornell and Larcker criteria

0.958

15

10 S. D. Chhetri et al.

Learning organization

0.585

0.477

Innovation

Trading partner

0.517

IT expertise

0.267

0.760

IT infrastructure

Top management support

0.132

HRA adoption

0.721

0.508

Financial readiness

0.176

0.520

External pressure

Relative advantage

0.568

Employee orientation

Readiness

0.177

0.128

Complexity

0.141

Compatibility

Awareness

1

0.153

0.336

0.554

0.362

0.263

0.445

0.170

0.251

0.260

0.211

0.350

0.686

0.473

2

0.381

0.316

0.542

0.210

0.219

0.278

0.189

0.183

0.381

0.225

0.383

0.350

3

0.631

0.395

0.517

0.255

0.382

0.435

0.166

0.085

0.612

0.563

0.558

4

0.637

0.614

0.145

0.464

0.498

0.327

0.588

0.195

0.460

0.677

5

0.555

0.593

0.179

0.505

0.578

0.374

0.501

0.168

0.734

6

0.711

0.309

0.318

0.527

0.402

0.593

0.456

0.256

7

0.110

0.258

0.122

0.321

0.169

0.381

0.181

8

0.355

0.336

0.118

0.670

0.562

0.460

9

Table 4. Discriminant validity-HTMT ratio

0.424

0.432

0.228

0.776

0.755

10

0.377

0.793

0.186

0.723

11

0.421

0.592

0.154

12

0.115

0.302

13

0.214

14

15

Validating the Measurement Scale Items on Readiness 11

12

S. D. Chhetri et al.

nations are still on initial phase of adoption of analytics. Hence, this study focus on validating the scale to measure readiness to adopt HR analytics in developing nations like Nepal. Reliability and Discriminant Validity was examined to check the reliability and validity. Out of 104 items, 78 items were established for the test. The study concludes that using this scale can further be validated and use to analyze the adoption level of HR analytics in the nations like Nepal. Awareness of HR analytics and its usage is still to be known by many organizations. There are numerous activities performed in HR department. Through research we can identify for what purpose analytics are used.

6 Research Limitations and Future Directions Although the study presents the reliability and validity test of the items for readiness to adopt HR analytics in the organizations of Nepal. There are number of shortcomings that should be considered when analysing the results. First, the survey is done to a small size and is limited to organizations of Nepal. However, more research must be done with larger size and in other nations. Second, the study has not considered moderating factors for adoption of HR analytics. Understanding whether other moderating factors can affect or impact the intention to adopt HRA for the purpose of changing adoption intention is necessary. Lastly, the study is a pilot survey done in Pokhara city of Nepal. Further research need to be carried out taking different cities of Nepal for validating the scale.

References 1. McCartney, S., Fu, N.: Bridging the gap: why, how and when HR analytics can impact organizational performance. Manag. Decis. 60(13), 25–47 (2021). https://doi.org/10.1108/ MD-12-2020-1581 2. Klimoski, R., et al.: Use of Workforce Analytics for Competitive Advantage, pp. 1–38. SHRM Report (2016) 3. Sesil, J.C.: Applying Advanced Analytics to HR Management Decisions: Methods for Selection, Developing Incentives, and Improving Collaboration (2013). [Online]. Available: https:// books.google.com/books?id=8R37AAAAQBAJ&pgis=1 4. Israrul Haque, M.: Human Resource Analytics: A Strategic Approach 5. Karma´nska, A.: The benefits of HR analytics. Pr. Nauk. Uniw. Ekon. Wroc. 64(8), 30–39 (2020). https://doi.org/10.15611/pn.2020.8.03 6. Kapoor, B., Kabra, Y.: Current and future trends in human resources analytics adoption. J. Cases Inf. Technol. 16(1), 50–59 (2014). https://doi.org/10.4018/jcit.2014010105 7. Alamelu, R., Nalini, R., Cresenta Shakila Motha, L., Amudha, R., Bowiya, S.: Adoption factors impacting human resource analytics among employees (2017). [Online]. Available: https://nsuworks.nova.edu/hsbe_etd 8. Johnson, R.D., Gueutal, H.G.: SHRM Foundation’s Effective Practice Guidelines Series the Use of E-HR and HRIS in Organizations Transforming HR Through Technology (2017). [Online]. Available: www.shrm.org/foundation 9. Akhmetova, S.G., Nevskaya, L.V.: HR Analytics: Challenges and Opportunities in Russian Companies. New Silk Road: Business …. atlantis-press.com (2020). https://doi.org/10.2991/ aebmr.k.200324.011

Validating the Measurement Scale Items on Readiness

13

10. van den Heuvel, S., Bondarouk, T.: The rise (and fall?) of HR analytics: a study into the future application, value, structure, and system support. J. Organ. Eff. 4(2), 157–178 (2017). https:// doi.org/10.1108/JOEPP-03-2017-0022 11. Wandhe, P.: HR analytics: a tool for strategic approach to HR productivity. SSRN Electron. J. (2020). https://doi.org/10.2139/ssrn.3700502 12. Bakre, D.M.P.: The role of HR analytics in the global village. Int. J. Trend Sci. Res. Dev. 3(2), 210.212.169.38 (2019) 13. Aghimien, D.O., Ikuabe, M., Aigbavboa, C., Oke, A., Shirinda, W.: Unravelling the factors influencing construction organisations’ intention to adopt big data analytics in South Africa. Constr. Econ. Build. 21(3), 262–281 (2021). https://doi.org/10.5130/AJCEB.V21I3.7634 14. Agrawal, K.P.: Investigating the determinants of big data analytics (BDA) adoption in Asian emerging economies. In: 2015 America’s Conference on Information Systems AMCIS 2015, pp. 1–18 (2015).https://doi.org/10.5465/ambpp.2015.11290abstract 15. Verma, S., Chaurasia, S.: Understanding the determinants of big data analytics adoption. Inf. Resour. Manag. J. 32(3), 1–26 (2019). https://doi.org/10.4018/IRMJ.2019070101 16. Ekka, S., Singh, P.: Predicting HR professionals’ adoption of HR analytics: an extension of UTAUT model. Organizacija 55(1), 77–93 (2022). https://doi.org/10.2478/orga-2022-0006 17. Nam, D., Lee, J., Lee, H.: Business analytics adoption process: an innovation diffusion perspective. Int. J. Inf. Manage. 49, 411–423 (2019). https://doi.org/10.1016/j.ijinfomgt.2019. 07.017 18. Ã, S.M.F., El-Masri, M.M.: Focus on research methods handling missing data in self-report measures, pp. 488–495 (2005). https://doi.org/10.1002/nur 19. Pilot Studies - Document - Gale Academic OneFile. [Online]. Available: https://go.gale. com/ps/i.do?p=AONE&u=googlescholar&id=GALE%7CA192589717&v=2.1&it=r&sid= AONE&asid=87470d30 20. Hair, J.F., Hult, G.T.M., Ringle, C.M., Sarstedt, M.: A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM), p. 165. Thousand Oaks, Sage (2017) 21. Ab Hamid, M.R., Sami, W., Mohmad Sidek, M.H.: Discriminant validity assessment: use of Fornell & Larcker criterion versus HTMT criterion. J. Phys. Conf. Ser. 890(1) (2017). https:// doi.org/10.1088/1742-6596/890/1/012163 22. Al-Skaf, S., Youssef, E., Habes, M., Alhumaid, K., Salloum, S.A.: The acceptance of social media sites: an empirical study using PLS-SEM and ML approaches. Adv. Intell. Syst. Comput. 1339, 548–558 (2021). https://doi.org/10.1007/978-3-030-69717-4_52

Optimizing Fire Response Unit Location for Urban-Rural Area Sunarin Chanta(B) and Ornurai Sangsawang King Mongkut’s University of Technology North Bangkok, 129 M.21 Noenhom, Muang Prachinburi 25230, Thailand {sunarin.s,ornurai.s}@itm.kmutnb.ac.th

Abstract. Fire safety is very important. Efficient facility planning is required to protect against destruction and reduce the risk of damage to facilities, and loss of lives, In this study, we propose an urban-rural maximal covering location model with the configuration of different types of areas for optimizing fire response units. Since the basic covering location model tends to cover more in a high population density area, with a limitation of resources a low population density area may leave as an unserved area. The objective is to maximize the demand that can be covered in standard response time. The GIS is utilized for managing the data and classifying the service area. A real-world case study is presented. The results show that solving problems with the proposed model can achieve full coverage with less number of facilities. Keywords: Facility location · Optimization · Rural · GIS · Fire department

1 Introduction The Royal Thai Government has launched the 20-Year Strategy for Thailand to achieve high-income status by 2036, namely Thailand 4.0. The government initiates the Eastern Economic Corridor (EEC) project aiming to revitalize the well-known eastern seaboard. The EEC is a special economic zone that consists of 3 provinces in the east of Thailand, namely Rayong, Chonburi, and Chachoengsao. The EEC development plan envisages a significant transformation of both physical and social development and plays an important role as a regulatory sandbox uplifting the country’s competitiveness. The objective of EEC development is to support the investment in the supercluster industry and the country’s target industries to be a mechanism to drive the economy in the next 20 years. The government has invested in the transportation infrastructure to support the industrial sector for enhancing the quality and connectivity within the country and the region more by land, sea and air, as well as to support the transportation and logistics hub of Asia in the future. The transportation infrastructure includes the construction of high-speed trains, dual-track trains, Laem Chabang Port, Sattahip Port, and U-Tapao Airport. Chonburi, one of the 3 provinces under the EEC project, is the province that has expanded the most representing 3.80% per year, followed by Rayong representing 3.17% © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 14–21, 2024. https://doi.org/10.1007/978-3-031-50158-6_2

Optimizing Fire Response Unit Location for Urban-Rural Area

15

per year, and Chachoengsao representing 1.57% per year, respectively. This data indicates the potential of the area where entrepreneurs have been setting up establishments continuously in the past 3 years and are still likely to expand more in the future. The information shows that Chonburi province tends to increase in terms of establishments, which is possible that accidents or risks in terms of fire may have a high chance of occurring. So in this case protection and prevention against damage from emergency incidents are essential. The Department of Disaster Prevention and Mitigation has a duty to prevent and extinguish the fire and various public disasters, including natural disasters and other emergency incidents. In this study, we aim to determine the optimal location of fire response units, which include fire trucks, staff, and necessary equipment for fire response units in the area of Chonburi province, Thailand. Since the area consists of both urban and rural land, so we proposed a mathematical model to solve the problem. The objective is to maximize the coverage in the whole area under the limitation of resources.

2 Literature Review The first mathematical model for facility location was the Maximal Covering Location Model (MCLP), which was first introduced by Church and ReVelle [1], with the objective of maximizing demands that can be covered subject to the limitation of the number of facilities. Demand can be covered when there exists a facility located within the standard response time. The MCLP model is well known and was applied for locating public facilities such as fire stations, ambulance stations, relief centers, etc. [2–6]. In the MCLP model, the standard response time is assumed to be the same for all the demands in the whole service area. However, in reality, there are some cases in which we cannot provide the same service standard for the whole area. In this study, we focus on the service area that has different population densities: urban and rural. Urban regions can be described as city areas with high population density, number of housing units, number of buildings, business areas, etc. On the opposite, rural regions are countryside areas with less population density [7]. More factors to classify areas such as age structure, registered resident structure, income level, etc. [8]. Therefore, with a limited number of facilities, public service departments tend to manage their facilities to be located in urban areas to save more people. The previous works related to facility location planning are briefly reviewed here. Chanta et al. [9] improved the quality of EMS in rural areas by providing a biobjective covering location model to locate ambulances at appropriate stations in the urban-rural area. The first objective was to maximize the expected number of requested calls that can be covered, while the second objective was to reduce disparities in service between different demographics. Three alternatives for the second objective had been proposed, which were minimizing the maximum distance between uncovered demand zones and their closest opened stations, minimizing the number of uncovered rural demand zones, and minimizing the number of uncovered demand zones. Karim et al. [10] integrated Geographic Information System (GIS)-based network analysis and multicriteria decision analysis (MCDA) technique for locating public service facilities in the districts of Buraidahthe city in the Kingdom of Saudi Arabia. The GIS

16

S. Chanta and O. Sangsawang

Network Analysis tools were applied to analyze the service area. Different coverages were set for different services including universities, hospitals, government services, etc. The Analytic Hierarchy Process (AHP) technique was used to determine the criteria weights, and the location-allocation model was used for suggestions of new service locations. Luo et al. [11] proposed a multi-objective optimization model to reduce urban-rural inequalities in EMS accessibility and coverage in Wuhan, China. The first objective was to minimize the total weighted distance of the uncovered demand from the nearest open EMS station in a rural area, which virtually maximized the accessibility of uncovered rural demand. The other two objectives were maximizing the service coverage and minimizing urban-rural inequality in service coverage. Liu et al. [12] proposed a location model for selecting the location of Park-and-Ride (P&R) facilities in the outer suburbs and in the urban areas. They considered the coverage demand characteristics, with the goal of maximizing the mileage of truncated private vehicles (for outer suburbs) and maximizing the demand for P&R facility coverage (for urban areas). Their P&R facility location model was constructed based on the travel choice behavior model and the progressive cooperation coverage model. For more review on service facility location, see Farahani et al. [13].

3 The Proposed Mathematical Model The proposed mathematical model is developed based on the formulations of the MCLP. Instead of using the same coverage, in this model the coverage is defined according to the types of areas. The urban areas require a short standard response time than the rural areas. The proposed mathematical model is described as follows. Index and Notations i = set of zones = 1, 2, …, n j = set of facility locations = 1, 2, …, m n = number of zones m = number of facility locations P = number of facilities to be located Parameters hi = number of calls in zone i r i = 1 if zone i is rural = 0 otherwise d ij = distance from zone i to facility location at j Du = coverage distance for urban area (the maximum distance that facility can respond to a call within the urban standard time) Dr = coverage distance for rural area (the maximum distance that facility can respond to a call within the rural standard time) aij = 1 if a call in zone i can serve by facility j within urban standard response time Du (d ij ≤ Du ) = 0 otherwise

Optimizing Fire Response Unit Location for Urban-Rural Area

17

bij = 1 if a call in zone i can serve by facility j within rural standard response time Dr (d ij ≤ Dr ) = 0 otherwise Decision Variables Y i u = 1 if urban zone i is covered by a facility = 0 otherwise Y i r = 1 if rural zone i is covered by a facility = 0 otherwise X j = 1 if there is a facility located at location j = 0 otherwise Optimization Model Maximize n 

hi ri Yir +

i=1

n 

hi (1 − ri )Yiu

(1)

aij Xj , ∀i

(2)

bij Xj , ∀i

(3)

i=1

Subject to Yiu ≤

m n   i=1 j=1

Yir ≤

m n   i=1 j=1 m 

Xj ≤ P

(4)

j=1

Xj ∈ {0, 1}, ∀j

(5)

Yiu , Yir ∈ {0, 1}, ∀i

(6)

The objective in Eq. (1) is to maximize the total expected calls that can be covered in a standard time. Note that in this study, there are two types of service areas, namely urban (with high a volume of calls) and rural (with a low volume of calls). So, the coverage conditions in the urban and rural areas will be different with respect to the area type. Constraints in Eqs. (2)–(3) force that a call can be covered if there exists a facility at a facility location. Note that a call in an urban zone can be covered within the urban standard response time, while a call in a rural zone may be covered within the rural or urban standard response time, respectively. Constraint in Eq. (4) limits the number of facilities to be located. Constraints in Eqs. (5)–(6) assign the domain of the decision variables.

18

S. Chanta and O. Sangsawang

4 A Case Study of Fire Department, Chonburi, Thailand Chonburi province has an area of approximately 4,363 km2 . The administrative area is divided into 11 districts and 92 sub-districts. The districts in Chonburi province are as follows: (1) Mueang, (2) Ban Bueng, (3) Nong Yai, (4) Bang Lamung, (5) Phan Thong, (6) Phanat Nikhom, (7) Sri Racha, (8) Koh Si Chang, (9) Sattahip, (10) Bo Thong, (11) Koh Jan. The total population is 1,566,875 people (as of 2020), with the density of 359.13 people per square kilometer. From the data of emergency notification via number 1669 in 2017, there were a total of 14,274 emergency calls, divided into 1,704 critical calls, 9,718 emergency calls, 2,539 non-urgent calls, 282 general calls, and 31 other calls. The data are analyzed based on the GIS software. The distribution of emergency calls is classified by type of emergency or severity shown in Fig. 1. As you can see from Fig. 1 Chonburi consists of both urban and rural areas. The high volume of emergency requested calls appeared more on side of the region. To manage the area and the calls, we divided the area of the whole province of Chonburi into 3 × 3 km2 grid size zones. A total of 609 zones were considered to be cut off since the zones did not have the incidents. Therefore, the total service area is 423 zones. The density of emergency incidents in each zone is shown in Fig. 2.

Fig. 1. Distribution of emergency calls in the area of Chonburi province, Thailand.

Optimizing Fire Response Unit Location for Urban-Rural Area

19

Fig. 2. Density of emergency calls in the area of Chonburi province, Thailand.

5 Computational Results To determine the proper location of the fire response units, we used the data of emergency calls via number 1669 in the year 2017 obtained from the National Institute of Emergency Medicine (NIEM). In this case, we have n = number of zones = 423, m = number of facility locations = 423. The area types are classified as rural (with a number of calls than or equal to 30 calls) and urban (with a number of calls more than 30 calls). The urban standard response time is set at 10 min, while the rural standard response time is set at 20 min. The number of facilities varies from 5 to 20. To compare the results of MCLP and the proposed model, we conduct 2 computational experiments. At first, we solve the problem with the MCLP model. The results of MCLP are shown in Table 1. Then, we solve the problem with the proposed model, where urban and rural areas are treated differently. The results of the proposed model are shown in Table 2. As you can, see that by using the proposed model, less number of facilities require for the same coverage. In this case, instead of leaving the rural areas uncovered due to the limitation of the number of facilities, we can serve the rural area with longer coverage.

20

S. Chanta and O. Sangsawang Table 1. The results of MCLP model

Number of units

Optimal location

Coverage

5

{68 81 103 179 270}

84.09

6

{68 81 103 179 270 295}

88.04

7

{34 69 81 103 179 270 295}

91.69

8

{34 69 76 110 180 186 251 313}

94.07

9

{34 69 76 110 180 186 251 313 334}

95.61

10

{34 69 76 110 180 186 251 295 334 364}

96.83

15

{26 38 43 60 110 146 151 180 244 251 260 312 334 375 389}

99.47

20

{10 26 29 34 84 96 125 154 182 186 232 246 279 289 333 343 383 396 403 418}

100.00

Table 2. The results of MCLP with urban-rural area model Number of units

Optimal location

Coverage

5

{69 104 111 198 251}

6

{68 81 103 179 270 295}

90.50

7

{34 69 81 103 179 270 295}

94.14

8

{34 69 76 110 180 186 251 313}

96.55

86.61

9

{34 69 76 110 180 186 251 313 334}

97.99

10

{34 69 76 110 180 186 251 295 334 364}

98.91

14

{13 18 25 99 111 153 162 189 232 277 287 346 353 407}

100.00

6 Conclusion and Future Research In this study, we purpose a mathematical model for determining the optimal fire response units for fire departments. A case study of Chonburi province, Thailand is presented. The objective is to maximize the number of incidents that can be responded to within the standard time. Since the case study area consists of urban and rural parts, so 2 different standard times are defined for both urban and rural areas. The GIS is applied to partition the service zone and classify the incidents over the response area. The results show that by using the proposed mathematical model, we efficiently optimal locate fire response units by maximizing the coverage under a low number of facilities. In this present study, we classify the service area based on incident density, and all incidents are treated the same. However, the severity of the incidents may different and requires different equipment and different types of rescue trucks. For future research, we would like to develop a model that considers different types of incidents corresponding to different types of facilities.

Optimizing Fire Response Unit Location for Urban-Rural Area

21

References 1. Church, R., ReVelle, C.: The maximal covering location problem. Pap. Reg. Sci. Assoc. 32, 101–118 (1974) 2. Pirkkul, H., Schilling, D.: The maximal covering location problem with capacities on total workload. Manage. Sci. 37(2), 233–248 (1991) 3. Black, B., Meamon, B.: Facility location in humanitarian relief. Int. J. Log. Res. Appl. 11(2), 101–121 (2008) 4. Chanta, S., Sangsawang, O.: Shelter-site selection during flood disaster. Lect. Notes Manag. Sci. 4, 282–288 (2012) 5. Alzahrani, A., Hanbali, A.: Maximum coverage location model for fire stations with top corporate risk locations. Int. J. Ind. Eng. Oper. Manag. 3(2), 58–74 (2021) 6. Srianan, T., Sangsawang, O.: Path-relinking for fire station location. In: Vasant, P., Zelinka, I., Weber, G.W. (eds.) Intelligent Computing & Optimization. ICO 2018. Advances in Intelligent Systems and Computing, vol. 866. Springer, Cham (2019) 7. U.S. Department of Agriculture, Economic Research Service, Rural Classification. Accessible at https://www.ers.usda.gov 8. Wang, J., Zhou, J.: Spatial evaluation of the accessibility of public service facilities in Shanghai: a community differentiation perspective. PLoS ONE 17(5), e0268862 (2022) 9. Chanta, S., Mayorga, M., McLay, L.: Improving emergency service in rural areas: a biobjective covering location model for EMS systems. Ann. Oper. Res. 221, 133–159 (2014) 10. Karim, A., Awawdeh, M.: Integrating GIS accessibility and location-allocation models with multicriteria decision analysis for evaluating quality of life in Buraidah City, KSA. Sustainability 12, 1412 (2020) 11. Luo, W., Yao, J., Mitchell, R., Zhang, X., Li, W.: Locating emergency services to reduce urban-rural inequalities. Socio-Econ. Plann. Sci. 84, 101416 (2022) 12. Liu, H., Li, Y., Hou, B., Zhao, S.: Optimizing the location of park-and-ride facilities in suburban and urban areas considering the characteristics of coverage requirements. Sustainability 14, 1502 (2022) 13. Farahani, R., Fallah, S., Ruiz, R., Hosseini, S., Asgari, N.: OR models in urban service facility location: a critical review of applications and future developments. Eur. J. Oper. Res. 276, 1–27 (2019)

Effects of Financial Literacy on Financial Inclusion: Evidence from Nepal’s Gandaki Province Deepesh Ranabhat1(B) , Narinder Verma1 , Pradeep Sapkota2 and Shanti Devi Chhetri2

,

1 Faculty of Management Sciences, Shoolini University, Bajhol, (HP) 173229, India

[email protected] 2 Faculty of Management Studies, Pokhara University, Pokhara-30, Kaski, Nepal

Abstract. Financial inclusion is regarded as an essential instrument for socioeconomic development. It has become a big issue in developing nations, as many people are still economically excluded. Different factors including financial literacy significantly influence financial inclusion. This study aims to determine how financial literacy impacts financial inclusion. The study area is the province of Gandaki in Nepal, from which a sample of one thousand respondents is selected for data collecting. For the study of independent and dependent variables, the Partial Least Squares Structural Equation Modelling method is applied. The examination of the data indicates that financial literacy considerably improves financial inclusion in the study region. This study concludes that financial inclusion can be improved by educating individuals on the significance of financial products and services. These findings aid policymakers in implementing financial literacy programs to improve financial inclusion. Keywords: Financial inclusion · Financial literacy · Development · Structural equation modelling · Nepal

1 Introduction Financial inclusion refers to the delivery of various financial services including deposits, loans, insurance, and remittances to everyone in the country easily and at a lower price. It is broadly considered an important tool for socio-economic development; however, financial inclusion is hampered by a lack of financial literacy. People will not demand financial products and services when they are unware about them. Functioning financial markets need knowledgeable customers, i.e. customers who are more financially literate to make better financial decisions. Literate customers demand more sophisticated financial services which enhance financial inclusion. Financial inclusion is affected by a person’s financial literacy level [1]. Financial literacy also improves the overall wellbeing of people by giving them the fundamental tools for budgeting, motivating them to save, and ensuring that they can live a respectable life after retirement [2]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 22–32, 2024. https://doi.org/10.1007/978-3-031-50158-6_3

Effects of Financial Literacy on Financial Inclusion

23

Financial literacy is a key concept in the context of financial inclusion. It is an ability to handle one’s financial resources successfully to achieve long-term financial security [3]. It is defined as an understanding of fundamental economic principles, along with the capacity to apply that knowledge and skill for the efficient use of money for long-term economic well-being [4]. According to the financial literacy-based approach of financial inclusion, increasing individuals’ financial literacy through education is necessary to achieve financial inclusion. If people are financially educated, they will be more likely to participate in the formal financial sphere. Financial literacy helps to inform people about the financial services and products available to them and their importance, which increases their participation in the formal banking services by having a bank account. Likewise, by having better financial literacy, people can benefit from additional advantages offered by the formal financial system, like investment and mortgage products. Financial literacy also helps individuals be independent and stable in their personal finances by educating individuals to discriminate between needs and wants, make and maintain a budget, save for timely payment of bills, and make retirement plans [5]. Financial inclusion has emerged as a significant concern in developing countries like Nepal, where a large proportion of the population (approximately 55%) still lacks accessibility of formal financial services [6]. People’s ability to access formal banking services is hampered by their low literacy level. In Nepal, 40% of adults are illiterate and the majority of the literate population have basic and secondary-level education only [7]. Because of low literacy levels, many individuals are not conscious of the financial services provided by the financial institution and their function. Due to significance of financial literacy, the central bank of Nepal, Nepal Rastra Bank (NRB), with an emphasis on the Government of Nepal (GoN), has set an objective to increase financial access through enhancing financial literacy and financial education. Different policies and programs including financial literacy programs have been conducted by the central bank of Nepal and other organizations to increase financial inclusion in Nepal. However, to the best of the researchers’ understanding, no study has been done in Nepal to determine how financial literacy impacts financial inclusion. Determining how financial literacy affects financial inclusion is crucial. Thus, this study seeks to quantify how financial inclusion is influenced by financial literacy in Gandaki province, Nepal.

2 Literature Review 2.1 Concept of Financial Literacy and Financial Inclusion Financial literacy is defined as understanding and use of financial knowledge, skills, and theories to help people make wise financial decisions for the betterment of their life [8]. It includes financial knowledge, awareness, attitude, skill, and behavior required to make wise economic decisions and attain the economic well-being of an individual [9]. Both definitions focus on understanding and use of financial concepts to obtain economic well-being. In this research, financial literacy refers to the combination of financial knowledge, skill, attitude, and behavior that help an individual to select and use various financial products and services for better economic life.

24

D. Ranabhat et al.

Financial inclusion as defined by Centre for Financial Inclusion is the state to which each person who is eligible to use financial services has accessibility of full range of superior services that are provided conveniently, at reasonable prices, and with consideration for the customers. CRISIL define financial inclusion as the degree of accessibility of formal financial service, comprising of services such as deposits, credits, remittance, insurance, and pensions, available to all sectors of society. [10] used three dimensions to construct the financial inclusion index. These dimensions comprise of accessibility, availability, and usage, all of which contribute towards the overall measurement of financial inclusion. Different researchers [11, 12], and [13] have used similar dimensions used by [10]. The term financial inclusion in this research pertains to the degree of access, availability, and utilization of financial services across all segments of the community. 2.2 Financial Literacy and Financial Inclusion Relationship Having financial literacy is a fundamental requirement for achieving financial inclusion [14]. Raising understanding of various financial services, saving, capital management, and credit management improves financial inclusion by increasing the demand for financial products [15]. Numerous studies have proven the connection between financial literacy and financial inclusion. A study by [16] discovered that financial literacy had a strong favorable influence on financial inclusion among 400 low-income rural Ugandans. The study also revealed that financial literacy enables low-income individuals to comprehend various banking services and make prudent judgments in order to maximize their use. [1] and [17] also discovered a statistically significant positive association between financial literacy and all metrics of financial inclusion. Similarly, according to [18], financial literacy positively promotes both savings and financial inclusion. The following research model (Fig. 1) and hypothesis (H1) have been created based on past research.

Fig. 1. Research model

Research Hypothesis (H1): Financial literacy significantly enhances financial inclusion.

Effects of Financial Literacy on Financial Inclusion

25

3 Research Methodology This study follows the positivism philosophy and is cross-sectional. It was conducted in Gandaki province, Nepal. The total households of 579,942 of Gandaki province [19] were used as population units and out of which 1000 households were chosen as a sample. A researcher-administered questionnaire with a fixed set of questions was developed to collect the necessary data which comprised of a socio-demographic profile along with 48 items (taken from various sources) on a five-point Likert scale to assess financial literacy and financial inclusion across four and three dimensions respectively. Both financial literacy and financial inclusion were regarded as higher-order constructs comprising various dimensions. The higher-order constructs with their dimensions are given in Table 1. In this study, the researchers followed multistage sampling for the collection of data. First, three districts were selected randomly out of 11 districts in the study area. Then the number of households to be taken from each district was calculated proportionately. Finally, the households from each district were selected using the snowball sampling (through referrals) and the required data were collected from the chief decision maker of selected household. The researchers visited the respondents and requested them to fill up the questionnaire. However, for illiterate respondents, the researchers took an interview and filled up the questionnaire. Data were collected for six months starting from April 2022. The researchers used SPSS and Smart-PLS for data analysis. Descriptive statistics such as frequency distribution in SPSS were used to assess the characteristics of the respondents, and Partial Least Squares Structural Equation Modelling (PLS-SEM) in Smart-PLS was applied to quantify the effect of financial literacy on financial inclusion.

4 Results and Analysis 4.1 Socio-demographic Summary This includes socio-demographic characteristics such as respondents’ district, gender, age, marital status, family type, caste, educational status, income per month, and expenditure per month. These variables were described through percentage frequency distribution using SPSS. For this study, participants were selected in proportion from three different districts. More than half (55.7%) of respondents are from the Kaski district. It is followed by 29% from the Syangja district and 15.3% from the Parbat district. The majority of households (64.5%) have a male as the primary decision maker, while only 35.5% of households have a female as the primary decision maker. Likewise, most of the respondents (87.5%) are married, majority of them are 41–50 years old (35.0%), live in a nuclear family (68.5%), and are Brahmin (53.4%). Similarly, the majority of them (44.3%) have a secondary education, have income of NRs. 10,001 to Rs. 40,000 per month (49.2%), and have expenditure of NRs. 20,001 to Rs. 40,000 per month (40.2%). 4.2 Structural Equation Modelling In this research, PLS-SEM was applied, which involves the measurement model and the structural model.

26

D. Ranabhat et al. Table 1. Constructs and their Indicators

Higher order construct

Lower order construct

Items

Indicators

Financial literacy

Financial knowledge

FinKw1, FinKw2, FinKw3, FinKw4, FinKw5, FinKw6, FinKw7, FinKw8, FinKw9

Knowledge about savings, loans, mortgages, cost of financial services, ATMs, type of insurance, remittance process, digital financial services, financial risk

Financial skill

FinS1, FinS2, FinS3, FinS4, FinS5, FinS6, FinS7

Ability to choose an appropriate account, fill up account opening form, determine cost and benefits from financial dealing, complete loan form, compute interest, fill remittance form, use ATM

Financial attitude

FinA1, FinA2, FinA3, FinA4, FinA5, FinA6

Readiness towards saving money, financial news, insurance, spending, loan from banks, using digital financial services

Financial behaviour

FinB1, FinB2, FinB3, FinB4

Habit of saving money, less spending, balancing income and expenses, maintaining financial records

Accessibility

FinAcc1, FinAcc2, FinAcc3, FinAcc4, FinAcc5, FinAcc6, FinAcc7

Convenient location, ease of access of employees, closeness of ATMs, convenient transaction time, access of information, no. of branches, fair account opening charges

Financial inclusion

(continued)

Measurement Model It applies reliability and validity tests to check the quality of the constructs. The researchers used a disjoint two-stage method to confirm reliability and validity of both higher-order constructs as well as their dimensions (lower-order constructs). First, the reliability and validity of dimensions were assessed and then of higher-order constructs.

Effects of Financial Literacy on Financial Inclusion

27

Table 1. (continued) Higher order construct

Lower order construct

Items

Indicators

Availability

FinAva1, FinAva2, FinAva3, FinAva4, FinAva5, FinAva6, FinAva7, FinAva8

Availability of various saving products, loan products, bank locker, ATM services, zero balance accounts, mobile/internet banking, fast service, assistance of banking staff

Usage

FinUse1, FinUse2, FinUse3, FinUse4, FinUse5, FinUse6, FinUse7

Frequency of deposit, frequency of withdraw, frequency of bank visit, frequency of using bank loan, payment of utilities, stock trading using bank, frequency of using locker facilities

Initially, the study used all 48 items of all dimensions (lower-order constructs) as given in Table 1. After deletion of 12 items (FinAcc5, FinAva1, FinAva2, FinAva5, FinUse2, FinUse4, FinK8, FinK9, FinS1, FinS7, FinA4, and FinA5), the reliability and validity of all dimensions were realized with 36 items. Then, the researchers checked the reliability and validity of higher-order constructs. For this, the researcher considered financial knowledge, skill, attitude, and behavior to be latent variables of financial literacy and usage, availability, and accessibility were treated as latent factors of financial inclusion, and further calculation were made using their scores. Validating Lower Order Constructs The researchers used Cronbach’s Alpha and Composite Reliability (CR) to determine reliability. Similarly, the construct validity includes convergent and discriminant validity. The researchers used Average Variance Extracted (AVE) for establishing convergent validity, and Heterotrait-Monotrait Ratio (HTMT), and Fornell and Larcker criteria for discriminant validity. Reliability and Convergent Validity The result of reliability test and AVE of lower-order constructs is presented in Table 2. All the values Cronbach’s alpha and CR tests (except Cronbach Alpha of financial attitude) are higher than the required minimum value of 0.70 [20] and the Cronbach Alpha of financial attitude is also more than 0.60, which is acceptable [21]. Hence, reliability is confirmed. Similarly, convergent validity is established when a latent unobserved construct has an AVE value of 0.50 or more [22]. The authors also mentioned that an AVE value of more than 0.40 is acceptable if the composite reliability of that construct is above 0.70. In the current study, all constructs (except financial attitude has AVE close to

28

D. Ranabhat et al.

0.50) are having an AVE value greater than the required limit of 0.50. Hence, convergent validity is established for all the constructs. Table 2. Reliability tests and AVE Construct

Cronbach’s alpha

CR

AVE

Access

0.863

0.898

0.594

Availability

0.833

0.882

0.601

Usage

0.763

0.84

0.513

Financial knowledge

0.888

0.912

0.597

Financial skill

0.881

0.913

0.678

Financial attitude

0.642

0.781

0.473

Financial behavior

0.754

0.837

0.563

Fornell and Larcker Criteria for Discriminant Validity Table 3 illustrates the square root of AVE for each construct (italicized diagonal values) and their corresponding correlation values. All of the square root of AVE values are higher than their respective correlation values, indicating strong evidence in favor of discriminant validity [22]. Table 3. Fornell and Larcker criteria for discriminant validity Acc

Avail

FA

FB

FK

FS

Access (Acc)

0.771

Availability (Avail)

0.505

0.775

Fin attitude (FA)

0.315

0.418

0.688

Fin behavior (FB)

0.267

0.308

0.531

0.751

Fin knowledge (FK)

0.327

0.304

0.424

0.378

0.773

Fin skill (FS)

0.309

0.335

0.442

0.331

0.71

0.824

Usage (Use)

0.373

0.227

0.369

0.343

0.474

0.443

Use

0.716

Discriminant Validity - Heterotrait-Monotrait Ratio (HTMT) Ratio Another broadly used technique to establish discriminant validity is HTMT. It measures the average correlation of indicators between the constructs. As shown in Table 4, all values are lower than the maximum threshold of 0.85, which confirms the existence of discriminant validity [23].

Effects of Financial Literacy on Financial Inclusion

29

Table 4. Discriminant validity-HTMT ratio Acc

Avail

FA

FB

FK

FS

Use

Access (Acc) Availability (Avail)

0.586

Fin attitude (FA)

0.395

0.545

Fin behavior (FB)

0.303

0.381

0.772

Fin knowledge (FK)

0.37

0.348

0.515

0.425

Fin skill (FS)

0.352

0.386

0.536

0.363

0.802

Usage (Use)

0.463

0.285

0.484

0.388

0.57

0.536

Higher Order Construct Validation After validating the lower-order construct, the validity and reliability of higher-order construct were examined. Table 5 shows that reliability was established using both Cronbach’s Alpha and CR. The Cronbach’s Alpha and CR values were found to be higher than 0.60 and 0.70, respectively, indicating that reliability was successfully established. Furthermore, an AVE greater than 0.50 indicates convergent validity (See Table 5). Likewise, Table 6 displays the outcomes of Fornell and Larcker criteria, indicating that the square root of AVE of financial inclusion and financial literacy (italicized diagonal values) is greater than their correlation. Moreover, Table 7 shows that the HTMT value is less than 0.85. Hence, it confirms the discriminant validity of financial literacy and financial inclusion. Table 5. Reliability tests and AVE—higher order constructs Cronbach’s alpha

CR

AVE

Financial inclusion

0.636

0.803

0.576

Financial literacy

0.779

0.858

0.603

Table 6. Discriminant validity of higher order constructs—fornell and larcker criteria Financial inclusion Financial inclusion

0.759

Financial literacy

0.608

Financial literacy 0.777

Structural Model Path analysis is used to investigate how financial literacy affects financial inclusion when the constructs’ validity and reliability are established. The outcomes of the path analysis are given in Table 8.

30

D. Ranabhat et al. Table 7. Discriminant validity of higher order constructs—HTMT method Financial inclusion

Financial inclusion Financial literacy

0.844

Table 8. Result of path analysis Association

Path coeff. (β)

T-stat

P-value

R2

Financial literacy → Financial inclusion

0.608

27.746

0

0.369

Table 8 reveals that financial literacy has statistically significant positive impact on financial inclusion (beta = 0.608, P-value < 0.001). And the coefficient of determination (R2 ) value is 0.369, which shows a moderate effect [20] in the established relationship. 4.3 Discussion This study was conducted to measure the effect of financial literacy on financial inclusion in Gandaki province, Nepal using PLS-SEM to quantify the effect. The measurement model in PLS-SEM found the constructs are reliable and valid and appropriate for conducting structural model. The finding shows that financial inclusion is significantly enhanced by financial literacy. The result is similar to those of [1, 16–18], indicating that financial education helps to increase financial inclusion. Financial literacy programs conducted by different organization enables individuals to understand various financial products and services and helps them to choose appropriate financial services. It also motivates individuals to use various financial services for better economic life. Thereby, financial literacy programs enhance financial inclusion. Thus, financial inclusion is highly influenced by financial literacy.

5 Conclusion In developing nations like Nepal, where many people remain financially excluded, financial inclusion has emerged as a key problem. Various policies and strategies, including financial literacy programs, have been devised and implemented to improve financial inclusion. This study discovered that financial literacy is a crucial factor in promoting financial inclusion. This study concludes that financial inclusion in Gandaki province can be enhanced by increasing financial literacy. Financial awareness programs raise understanding of different financial products and their significance and inspires individual to use them which promotes financial inclusion. The research’s findings complement the financial inclusion theory based on financial literacy and earlier literatures that describe the significance of financial literacy on financial inclusion. This research advises to the policymakers to implement financial literacy programs for individuals in order to expand financial inclusion.

Effects of Financial Literacy on Financial Inclusion

31

6 Research Limitations and Opportunities for Future Research In this study, Gandaki province was taken as a study area. In Nepal, due to diverse geographic areas and unequal access to financial services, the financial literacy level as well as financial inclusion is not same in all provinces. So, the results may not apply in other provinces of Nepal. In the future, a comparative study of different provinces can be conducted. Likewise, an experimental and comparative study can be conducted by providing financial literacy to one group and controlling another group and making a comparative study between two groups.

References 1. Grohmann, A., Klühs, T., Menkhoff, L.: Does financial literacy improve financial inclusion? cross country evidence. World Dev. 111(95), 84–96 (2018). https://doi.org/10.1016/j.wor lddev.2018.06.020 2. Ramachandran R.: Financial literacy-the demand side of financial inclusion. In: 26th SKOCH Summit 2011 Swabhiman-Inclusive Growth and Beyond 2 nd & 3 rd June, Mumbai India, no. ii, pp. 1–16 (2011) 3. Stolper, O.A., Walter, A.: Financial literacy, financial advice, and financial behavior. J. Bus. Econ. 87(5), 581–643 (2017). https://doi.org/10.1007/s11573-017-0853-9 4. Hung, A.A., Parker, A.M., Yoong, J.K.: Defining and measuring financial literacy (2009). [Online]. Available: https://www.rand.org/content/dam/rand/pubs/working_papers/ 2009/RAND_WR708.pdf 5. Ozili, P.K.: Munich personal RePEc archive theories of financial inclusion theories of financial inclusion. Munich Pers. RePEc Arch. (104257) (2020) 6. Demirgüç-kunt, A., Klapper, L., Singer, D., Ansar, S., Hess, J.: The global findex database. The World Bank, Washington (2017) 7. Nepal Rastra Bank.: Financial Inclusion Roadmap 2017–2022 (2020) 8. OCDE.: PISA 2012 Results: Students and Money: Financial Literacy Skills for the 21st Century, vol. VI (2014) 9. Nepal Rastra Bank. Financial literacy framework (2020) 10. Sarma, M.:Index of Financial Inclusion (2008) 11. Yadav, P., Sharma, A.K.: Financial inclusion in India: an application of TOPSIS. Humanomics 32(3), 328–351 (2016). https://doi.org/10.1108/H-09-2015-0061 12. Hanivan, H., Nasrudin, N.: A financial inclusion index for Indonesia. Bul. Ekon. Monet. dan Perbank. 22(3), 351–366 (2019). https://doi.org/10.21098/bemp.v22i3.1056 13. Ranabhat, D. Verma, N., Sapkota, P., Chhetri, S.D.: Impact of financial inclusion on social and economic well-being of households: a case of Kaski District, Nepal. In: BT—Intelligent Computing & Optimization, pp. 1027–1037 (2023) 14. Vishvesh, R., Venkatraman, B.: Financial inclusion through financial education. Int. J. Appl. Bus. Econ. Res. 13(1), 19–35 (2015) 15. Verma, S., Oum Kumari, R.: Role of financial literacy in achieving financial inclusion. Int. J. Appl. Bus. Econ. Res. 14(6), 4607–4613 (2016) 16. Bongomin, G.O.C., Munene, J.C., Ntayi, J.M., Malinga, C.A.: Nexus between financial literacy and financial inclusion: examining the moderating role of cognition from a developing country perspective. Int. J. Bank Mark. 36(7), 1190–1212 (2018). https://doi.org/10.1108/ IJBM-08-2017-0175 17. Ghosh, S.: Financial literacy and financial inclusion unbundling the nexus. Econ. Polit. Wkly. 54(13), 75–82 (2019)

32

D. Ranabhat et al.

18. Morgan, P.J., Long, T.Q.: Financial literacy, financial inclusion, and savings behavior in Laos. J. Asian Econ. 68(7) (2020). https://doi.org/10.1016/j.asieco.2020.101197 19. CBS Nepal.: Nepal Census 2021, pp. 1–96 (2022) [Online]. Available: https://census nepal.cbs.gov.np/Home/Details?tpid=5&dcid=3479c092-7749-4ba6-9369-45486cd67f30& tfsid=17 20. Hair, J.F., Ringle, C.M., Sarstedt, M.: PLS-SEM: Indeed a silver bullet. J. Mark. Theory Pract. 19(2), 139–151 (2011). https://doi.org/10.2753/MTP 21. Ursachi, G., Horodnic, I.A., Zait, A.: How reliable are measurement scales? external factors with indirect influence on reliability estimators. Procedia Econ. Financ. 20(15), 679–686 (2015). https://doi.org/10.1016/s2212-5671(15)00123-9 22. Fornell, C., Larcker, D.F.: Evaluating structural equation models with unobservable variables and measurement error. J. Mark. Res. 18(1), 39 (1981). https://doi.org/10.2307/3151312 23. Henseler, J., Ringle, C.M., Sarstedt, M.: A new criterion for assessing discriminant validity in variance-based structural equation modeling. J. Acad. Mark. Sci. 43(1), 115–135 (2014). https://doi.org/10.1007/s11747-014-0403-8

Business and Information Technology Strategy Impact on Organizational Performance: A Case Study of Nepal Telecom Sudip Poudel1 , Neesha Rajkarnikar1 , Deepanjal Shrestha1,2(B) Deepmala Shrestha1 , and Seung Ryul Jeong3

,

1 School of Business, Pokhara University, Pokhara, Nepal

[email protected], [email protected], [email protected], [email protected] 2 Nanjing University of Aeronautics and Astronautics, Nanjing, China 3 Graduate School of Business IT, Kookmin University, Seoul, South Korea [email protected]

Abstract. An effective business and information technology (IT) strategy has a significant impact on organizational performance. This study examines the impact of Nepal Telecom’s business and IT strategy on organizational performance. The study is based on primary data and employs descriptive and analytical research design to examine the impact of independent variables (business strategies, IT strategies, individual’s decision-making capacity, IT implementations, technological deployment and budget allocations) on the dependent variable (organizational performance of Nepal telecommunication. The respondents are employees of Nepal Telecommunications working in technical as well as managerial positions. The findings of this study are significant to various telecom companies to understand and identify business and IT strategies that are required to take decisions for improving organizational performance. The research provides information on the latest technologies and approaches that the company can use to improve its operations and service delivery and increase customer satisfaction. This study helps to understand how appropriate business and IT strategies can help to gain a competitive advantage and drive the performance of a business. This research can also help the company to make wise decisions on investments and resource optimization for better growth and better organizational performance. Keywords: IT strategy · Business strategy · Telecommunications · Services · Organizational performance · Nepal telecom

1 Introduction The telecommunications sector is rapidly evolving day by day with the help of the latest technological developments. The technological developments in the world require businesses to use information systems more than they used to. Establishing relationships between the various resources is important to meet the organizational goals and attain © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 33–44, 2024. https://doi.org/10.1007/978-3-031-50158-6_4

34

S. Poudel et al.

the fully desired organizational performance. According to Kaplan, for a competitive advantage to be created in an organization, its business and information technology strategies and processes must be aligned well. Although the strategic relationship has a history of being rated as a top issue facing business and IT executives, no single solution has been established to address this issue [1]. Organizations today operate in dynamic and constantly changing environments, requiring continuous assessment of strategic relations to ensure that they operate at an optimal level. A key success factor for a successful company to meet its performance requirements in a dynamic environment is effective and efficient information technology supporting business strategies and processes [2]. Combining the business strategy with the Information systems strategy has become a vital concern in the planning process of organizations [3]. Today, Information technology (IT) is a very important part of an organization’s daily operations and business strategy. IT is viewed as a technological resource that would help organizations does better things [4]. The consequences of not adopting effective strategic planning in an organization could lead to the inability of the organization to deal with the rapidly changing environment. When the pace of doing business is slow, managers could operate on the assumption that the future, would be substantially like the past and thus could establish plans/goals simply by extrapolating from past experiences. Unfortunately, today events are moving too rapidly for experience to be a reliable guide to a manager. Another effect of not adopting an effective strategy in an organization is the inability to have clearly defined objectives and methods for achieving these objectives. Thus, such organizations may not have a clear purpose and direction. It is important to study this area so that business and information technology strategies must align in such a way that organizations perform better in all aspects. 1.1 History of Nepal Telecom The start of telecommunication service in Nepal can be traced back to 1973, with the establishment of Mohan Akashwani in B.S. 2005 [5]. The Telecommunication Department was established in 2012–2017 (BS), and the Telecommunications Development Board was converted into Telecommunications Development Board in 2026 (BS). Nepal Telecommunications Corporation was established in 2032 (BS) and transformed into Nepal Doorsanchar Company Limited (NDCL) in 2061. Nepal Telecom has adopted and implemented changes in its business model in response to rapid advancements in technology, high expectations of customers, and ever-changing market situations. Today, Nepal Telecom has no monopoly on the market due to competition from other telecom players such as Ncell, UTL, Smart Telecom, Nepal Satellite Telecom Company Ltd, and around 50 Internet Service Providers (ISP). After Ncell was acquired by Telia Sonera group and Celcom Axiata group, the state-owned company has been facing fierce completion and many challenges and threats in the market in voice telephony as well as internet service [6]. The current market share of around 92% belongs to Nepal Telecom and Ncell, making it a duopoly irrespective of the presence of the other 3 telecom operators.

Business and Information Technology Strategy Impact

35

1.2 Problem Statement IT is an important part of an organization’s daily operations and business strategy. It is observed over the past three decades that there has been a rapid change in technology. Not adopting an effective IT-based strategy in an organization could lead to the inability to deal with the rapidly changing environment. Further, there is a proper need for IT strategy alignment with the business strategies so that business objectives can be achieved. Thus, it is important to study this area and fill in the gap by defining the variables that tend to influence IT strategies and business strategies in an organizational setup. This study describes Nepal Telecom’s business and IT strategy and its impact on organizational performance. The study has considered multiple types of factors that influence organizational performance that is decisive on various levels of organizational performance. These factors include business strategies, IT strategies, technological deployment, IT implementation, the impact of IT on individuals’ decision-making capacity, budget allocation, and organizational performance. 1.3 Hypothesis • H1: There is a significant impact of business strategies on the organizational performance of Nepal telecom. • H2: There is a significant impact of IT strategies on the organizational performance of Nepal telecom. • H3: There is a significant impact of individual decision-making capacity on the organizational performance of Nepal telecom. • H4: There is a significant impact of IT implementations on the organizational performance of Nepal telecom. • H5: There is a significant impact of technology deployment on the organizational performance of Nepal telecom. • H6: There is a significant impact of budget allocations on the organizational performance of Nepal telecom. • H7: Educational level of employees plays a moderating role in the research model concerning the impact of business and IT strategies on the organizational performance of Nepal telecom. • H8: Work experience of employees plays a moderating role in the research model concerning the impact of business and IT strategies on the organizational performance of Nepal Telecom 1.4 Significance of the Study In Nepal, there is a lack of prior research on how business and information technology strategies impact organizational performance. NTC is the biggest telecom company, and taking it as a case study will help the company be more competitive and demanding. This research helps NTC in identifying business and IT strategies to develop more effective and efficient decisions, as well as providing information on the latest technologies and approaches to improve operations and service delivery. The research is designed to help the company drive its performance and anticipate and respond to changes in the market and industry.

36

S. Poudel et al.

2 Literature Review The literature for both the global context and local context was considered to trace the earlier studies and find the gap in the literature. It was observed that a good number of studies are carried out at the global level in this sector and many aspects related to the basic theme are studied. However, there is a huge gap in the study of this aspect in the Nepalese context. It was required that the gap in this area in the context of Nepal must be brought to light which has helped this study to come up. Further, the growth of telecommunication and IT-based industries in Nepal is witnessing a sharp rise with many technology-based solutions and services offered by them [8]. The need to align the IT and business strategies is important to guarantee good organizational performance and meet the very objectives of the business (Table 1).

3 Research Methodology 3.1 Conceptual Framework The conceptual framework defines the independent variables, dependent variables and moderating variables of the study as shown in figure the business strategy, IT strategy, individual decision-making capacity, IT implementation, technological deployment and budget allocation are the independent variables that are studied for dependent variable organizational performance. The study also considers educational qualification and work experience as moderating variables (Fig. 1). 3.2 Data Collection, Analysis Method and Reliability The study is based on primary data and some secondary data to understand the underlying concepts and build a strong foundation for the study. Primary data is used to test the hypothesis and test the evidence of the claim. A survey questionnaire method is used for 40 respondents and the study has two parts. The first part includes the personal and general information of the respondents. The second part investigated the factors affecting the various independent variables that have impact on the overall organizational performance. The respondents of the study include business IT and business managers working in Nepal Telecom, head office in Pokhara and other cities of Nepal. Quantitative data analysis is used in the study to meet the specified research objectives. Both descriptive and inferential statistical methods of data analysis are used. Descriptive statistics is used to compute mean, standard deviation, frequency distributions of the data, which is presented in graphs, charts, and cross-tabulations etc. Also, different inferential tests are performed for the analysis like ANOVA test, T-test and correlation. The reliability test is performed using Cronbach’s α where all the variables are found to be greater than 0.7, indicating that the measurement data is reliable. Further, the overall reliability of the scale in the study is 0.758, conferring to the overall reliability of the questionnaire.

Business and Information Technology Strategy Impact

37

Table 1. Literature review with key theories and findings Ref

Year, Author

Findings

[9]

2021, Dairo M, Adekola J, Apostolopoulos C, Tsaramirsis G

Contribute to the practical understanding of strategic alignment and demonstrates the applicability and robustness of the Strategic Alignment Model (SAM). It also analyzes potential opportunities and risks associated with the IT strategy alignment with business

[10]

2003, Grover, V., & Teng, J. T. C

The study shows the functional integration as a horizontal relationship, and the extension of the concept of strategic fit to the functional domain of the strategic alignment framework

[11]

2017, Al-hamadi

This study discusses emerging trends in the telecommunications industry, including the growth of mobile data, the emergence of 5G technology, and the increasing use of artificial intelligence in the industry. It also explores the potential impact of these trends on the industry and its future outlook

[12]

2020, D. Shrestha, T. Wenan, SR Jeong

Understanding the attitude of tourist and tourism SMEs for ICT and digital technologies in the Nepalese context. The aspects are studied under TAM model

[13]

2003, Mintzberg

The study brings out the strategy process, concepts, contexts and cases in business. It highlights the various authors and managers and their use of concepts differently that might include other related terms such as goals and objectives as part of the strategy, whereas others make firm distinctions between them

[14]

2020, Deepanjal Shrestha, Tan Wenan, Adesh Khadka, and Seung Ryul Jeong

This study has examined the ICT and digital technology implementation at the Nepal government operational level. It provides insight on how the government is shaping policies for ICT and digital technologies. It also provides the different technologies in existence in Nepal and how these services are used in the current context (continued)

38

S. Poudel et al. Table 1. (continued)

Ref

Year, Author

Findings

[15]

1999, Wetherbe J, E. Turban, and E. Mclean

The study examines Information Technology for management and how making connections for strategic advantages are important in a technology-oriented business environment

[16]

2017, Sibanda, Mabutho and Ramrathan, Durrel

The study portrays the regulatory challenges in the telecommunications Industry. The study has built a comprehensive landscape of regulatory policies and how they impose challenges in its implementations

Independent Variables Business Strategy IT Strategy Individuals Decision Making Capacity IT Implementations Technological Deployment Budget Allocations

Dependent Variable Organizational Performance

Moderating variables Educational Qualification Work Experience

Fig. 1. Conceptual framework of the study

4 Data Analysis and Results 4.1 Respondents’ Profile The demography data of the respondents is depicted in Table 2 and work experience is shown in Table 3 followed by qualification in Table 4 and employee profile in Table 5. Table 2. Socio-demographic data Gender

Frequency

Percent

Male

15

37.5

Female

25

62.5

Total

40

100

Business and Information Technology Strategy Impact

39

Table 3. Distribution by work experience in NTC Work experience period

Frequency

Percent

Less than 2 years

3

7.5

2–5 years

16

40.0

5–10 years

11

27.5

Above 10 years

10

25.0

Total

40

100.0

Table 4. Qualification of employees Educational level

Frequency

Percent

Plus 2/diploma

6

15.0

Bachelors

19

47.5

Masters and above

15

37.5

Total

40

100.0

Table 5. Employee job profile Work departments

Frequency

Percent

Office of COO (Chief Operating Officer)

9

22.5

Office of CTO (Chief Technical Officer)

6

15.0

Office of CCO (Chief Commercial Officer)

4

10.0

Office of CFO (Chief Finance Officer)

6

15.0

Office of CHRO (Chief Human Resource Officer)

4

10.0

Wireline and CSD(Customer service department)

5

12.5

Wireless service directorate

4

10.0

Others

2

5.0

Total

40

100.0

4.2 Analysis of Independent Variables The data analysis depicts that the mean value of business strategy dimensions is more than 3 with a 0.918 SD, which indicates that overall respondents are pleased with the performance offered by Nepal Telecom. Similarly, the IT strategy dimensions also show a mean value of 3.48 with a 0.829 SD, which again confirms that there is a positive impact of IT strategy on the overall organizational performance. The mean value of 3.75 with a SD of 0.825 shows that the impact of IT on individual decision-making capacity has an influential role on overall organizational performance. Further, the mean value of

40

S. Poudel et al.

3.51 with an SD of 0.859 shows that technological deployment dimensions have a great contribution to organizational performance. In addition, the budget allocation dimensions also depict the mean value to be 3.25 with SD of 0.927, indicating satisfaction with the budget allocation and the impact of the budget on organizational performance. 4.3 One-Way ANOVA

Table 6. Impact of business and information technology strategies on the organizational performance on the basis of the educational level of respondents. Educational level

Mean

Std. deviation

Minimum

Maximum

Fvalue

Pvalue

Plus 2/Diploma

3.160

.907

1.7

4.25

0.504

0.608

Bachelors

3.293

.743

2.0

4.25

Masters

3.486

.677

1.7

4.25

Source: SPSS Output, 2022, data survey and interview

The impact of business and information technology was studied with respect to the education level, and a one-way ANOVA test was carried out on the collected data. It was observed that the educational level of respondents’ mean value was in the range of 1.7–4.25, with an average of 3.160 and a standard deviation (SD) of 0.907. Similarly, the bachelor’s degree holders had a value in the range of 2.0–4.25, with an average of 3.293 and a standard deviation of 0.743. Further, respondents with a master’s degree had a value ranging from 1.7 to 4.25, with an average of 3.486 and an SD of 0.677. Since p-value = 0.608 > level of significance (α) = 0.05, hypothesis H7 was rejected (Table 5). The same results can be drawn for Table 6, and it is seen that hypothesis 8 is also rejected (Table 7). Table 7. Impact of business and information technology strategies on organizational performance based on years of service. Years of working

Mean

Std. deviation

Minimum

Maximum

F value

P value

Less than 2 years

3.541

.921

2.50

4.25

0.109

0.954

2–5 years

3.294

.786

1.75

4.25

5–10 years

3.392

.812

1.71

4.25

Above 10 years

3.316

.603

2.25

4.00

Source: SPSS Output, 2022, data survey and interview

4.4 Inferential Statistical Analysis Correlation Analysis

Business and Information Technology Strategy Impact

41

Table 8. Correlation between business and technology strategies dimensional factors and Organizational performance of Nepal telecom

OP BS ITS IDM

OP

BS

ITS

IDM

ITM

TD

BA

1

.340*

.410**

−.085

.209

−.188

.253

.032

.009

.604

.195

.245

.116

1

.170

−.076

.177

.165

−.338

.293

.640

.274

.310

.033

1

.012

.250

.254

.065

.941

.120

.113

.688

1

.265

.000

−.087

.098

1.000

.592

1

.051

-.253

.752

.115

1

−.197

ITM TD

.224 BA

1

Source: SPSS Output, 2022, data survey and interview

Table 8 shows the correlation matrix between the independent variable and dependent variables. The level of significance (p = 0.032, p < 0.05) indicates that there is a significant impact of business strategy on organizational performance. The level of significance (p = 0.009 p < 0.05) indicates that there is a significant impact of IT strategies on organizational performance. The level of significance (p = 0.604, p > 0.05) indicates that there is no significant impact of an individual’s decision-making capacity on organizational performance inferring that there is no correlation between them. Further, the level of significance (p = 0.195, p > 0.05) indicates that there is no significant impact of IT on organizational performance. The Pearson coefficient value of 0.209 indicates that the two variables have a weak correlation and are positive. The level of significance (p = 0.245, p > 0.05) indicates technological deployment has no that there is no significant impact on organizational performance. The Pearson coefficient value of −0.188 indicates that these two variables are not correlated and have negative correlation. The level of significance (p = 0.116, p > 0.05) indicates that there is no significant impact of budget allocations on organizational performance. ANOVA analysis was conducted to determine the significance of the regression model. It can be seen that the p-value of the model is 0.001 which is less than the significance level of the study (i.e. p < 0.05) so, the hypothesis was accepted. This means there is a significant impact between the dependent variables and independent variables as shown in Table 9. In this study, the regression model was derived for reaching the point of conclusion as follows: OP = β0 + β1BS + β2ITS + β3IDM + β4ITM + β5TD + β6BA where,

(1)

42

S. Poudel et al.

OP = Organizational performance BS = Business Strategy ITS = IT strategy IDM = Individuals decision making ITM = IT implementations TD = Technological deployment BA = Budget allocation Table 9. ANOVA for multiple regression Model 1

df

Mean square

F

Sig

9.787

6

1.63

4.781

.001b

Residual

11.260

33

.341

Total

21.048

39

Regression

Sum of squares

Source: SPSS Output, 2022, data survey and interview

From Table 10, we can determine the regression model as follows: OP = −0.257 + 0.564BS + 0.415ITS − 0.82IDM + 0.167ITM − 0.321TD + 0.337BA (2) The hypothesis testing shows that H1 and H2 were accepted with 0.006 and 0.019 values, whereas H3 and H4 were rejected with 0.589 and 0.235 values. Further, hypotheses H5 and H6 were accepted with values 0.044 and 0.021. Table 10. Regression coefficient analysis of variables Coefficient analysis in Table

Beta

T-value

P value

Alternative hypothesis

(Constant)

−.257

−.203

0.84

Business strategy

.564

2.926

.006

Accepted

IT Strategy

.415

2.470

.019

Accepted

Individual decision making

−0.82

−0.545

.589

Rejected

IT implementations

0.167

1.211

.235

Rejected

Technological deployment

−0.321

−2.091

.044

Accepted

Budget allocations

0.337

2.427

.021

Accepted

R–square

.465

F

4.781

P value

0.001

Source: SPSS Output, 2022, data survey and interview

Business and Information Technology Strategy Impact

43

5 Conclusion The study on the impact of business and information technology strategies on organizational performance at Nepal Telecom concluded that business and IT strategies had a significant influence on organizational performance. The study found that individual decision-making capacity and technological deployment did not have a significant influence on organizational performance. Additionally, the moderating effect of educational level and work experience among different respondents of different departments and age groups and responsibilities was found to have no significant effect. The study concluded that a strong business strategy is an important component and must align with the company’s overall goals and objectives to increase organizational performance. Further, IT strategies are essential for the success of a telecom company, as they improve efficiency, reduce costs, and increase customer satisfaction. The IT strategies were also important to gain a competitive advantage in business and introduce new products and services.

References 1. Farida, I., Setiawan, D.: Business strategies and competitive advantage: the role of performance and innovation. J. Open Innov. Technol. Mark. Complex. 8, 163 (2022). https://doi. org/10.3390/joitmc8030163 2. Klein, J.T., Sorra, J.S.: The challenge of innovation implementation. Acad. Manag. Rev. 21(4), 1055–1080 (1996) 3. Hofstede, G.: A study of business and IT alignment in the 21st century. J. Manag. Inf. Syst. 20(2), 193–212 (2003) 4. Jonas, H., Thomas, K.: The business model concept: theoretical underpinnings and empirical illustrations. Eur. J. Inf. Syst. 12(1), 49–59 (2010). https://doi.org/10.1057/palgrave.ejis.300 0446 5. Nisha, M.: Financial analysis of public comany and its contribution to Nepalese economy: a case study of Nepal Telecom, Faculty of Management Tribhuvan University, (2020). https:// elibrary.tucl.edu.np/bitstream/123456789/9769/2/Proposal%20-NTC.pdf 6. Shrestha, D., Devkota, B., Jeong, S.R.: Challenges and factors affecting E-governance practices in Nepal. In: 9th International conference on software, knowledge, information management and applications (SKIMA), vol. 9. Kathmandu, Nepal 15–17 December (2015). https:// doi.org/10.1109/SKIMA.2015.7399981 7. Masume, P., Neda, A., Saeedeh, R.H.: The impact of IT resources and strategic alignment on organizational performance: the moderating role of environmental uncertainty. Digital Business 2(2), 100026 ISSN 2666–9544 (2022). https://doi.org/10.1016/j.digbus.2022. 100026 8. Wenan, T., Shrestha, D., Shrestha, D., Rajkarnikar, N., Jeong, S.R.: The role of emerging technologies in digital tourism business ecosystem model for Nepal. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol 569. Springer, Cham. https://doi.org/10.1007/978-3-031-19958-5_105 9. Dairo, M., Adekola, J., Apostolopoulos, C., Tsaramirsis, G.: Benchmarking strategic alignment of business and IT strategies: opportunities, risks, challenges and solutions. Int. J. Inf. Technol. 13(6), 2191–2197 (2021). https://doi.org/10.1007/s41870-021-00815-7 10. Grover, V., Teng, J.T.C.: Business/IT alignment: Achieving a durable partnership. J. Manag. Inf. Syst. 20(1), 191–206 (2003)

44

S. Poudel et al.

11. Al-Hamadi, H.M.: Emerging trends in the telecommunications industry: a review. J. Netw. Comput. Appl. 88, 1–10 (2017) 12. Wenan, T., Shrestha, D., Shrestha, D., Jeong, S.R.: Attitude of international tourist towards ICT and digital services in tourism industry of Nepal. In: Association for Computing Machinery (AISS 2019), pp. 1–7. USA, Article 10 (2019). https://doi.org/10.1145/3373477.3373487 13. Mintzberg, H., Lampel, J., Quinn, J.B., Ghoshal, S.: The strategy process: concepts, contexts, cases. Fourth Edition, Pearson Education Limited (2003) 14. Shrestha, D., Wenan, T., Khadka, A., Jeong, S.R.: Digital tourism security system for Nepal. KSII Trans. Internet Inf. Syst. 14(11), 4331–4354 (2020). https://doi.org/10.3837/tiis.2020. 11.005 15. Wetherbe, J., Turban, E., Mclean, E.: Information Technology for management: Making connections for strategic advantages, 2nd edn. NY, John Wiley and Sons Inc, New York (1999) 16. Tannenbaum, R.S.: Regulatory challenges in the telecommunications industry: a literature review. J. Manag. 36(2), 307–327 (2010)

A Systematic Literature Review on Factors Affecting Rural Tourism Pradeep Sapkota1(B)

, Kamal Kant Vashisth1

, and Deepesh Ranabhat2

1 Faculty of Management Sciences, Shoolini University, Bajhol, Himachal Pradesh 173229,

India [email protected] 2 Faculty of Management Studies, Pokhara University, Pokhara-30, Kaski, Nepal

Abstract. Studies related to rural tourism have been done from a different perspective, but lack of proper systematic reviews has been observed on factors affecting rural tourism. To overcome the problems, this paper has been conducted. This investigation reviewed articles related to factors affecting rural tourism found in the Scopus database from 2001 to 2021. The research subject, purpose, techniques, and data source of 61 articles from various journals were selected and listed in a complete table, showing the study topic, purpose, methods, and data source of articles. This article focuses on general characteristics, citation analysis, and keyword co-occurrence analysis by using the VOS viewer program. The results expose the trend and impact of literature, journals, and studied area on the related topic together. Hence, this study can unfold and boost the knowledge of the related topic and give an idea of a new research trend. Keywords: Tourists · Rural tourism · Systematic literature review · Citation analysis · Bibliometric analysis

1 Introduction Tourism in rural areas is regarded as an activity that can help to solve rural problems by forging strong links between tourism and rural life [1]. Tourism in rural areas in the past was observed only as a monetary compliment for farmers but now it is an important sector which contribute to generate social benefits as well [2]. Rural tourism “includes a wide range of leisure activities carried out in rural regions, including community-based tourism, ecotourism, cultural tourism, adventure tourism, guest farms, backpacking, horseback riding, homestays, and agritourism [3]. The purpose of visiting these places is to learn about, experience, and appreciate their unique cultures and natural features [4]. Rural tourism plays an important role in the economic development of the community and to spend better living standard. Most of such research mainly focuses on making rural tourism more successful and sustainable [5]. Market-oriented development has been proved in numerous studies to be a viable method for rural tourism. Similarly, stakeholder’s views play important role for the development of rural tourism [6]. Quantitative methods are commonly used to identify the key variables and factors influencing © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 45–55, 2024. https://doi.org/10.1007/978-3-031-50158-6_5

46

P. Sapkota et al.

stakeholders’ perceptions [7]. However, several research are focused on either the service provider or the tourist’s perspective which is not enough for the success of rural tourism. To comprehend the factors influencing rural tourism development, it is necessary to look at both the demand and supply sides of the tourism industry. Over the past two decades, it has been observed that there has been tremendous growth of academic research in the field of rural tourism. Most of the research related to rural tourism is focused on developed countries. Only a very few studies have been conducted in developing nations [8, 9]. As interest in the success of rural tourism, lot of research related to rural tourism have been conducted recently but a core study related to key success factors of rural tourism is yet to be done through systematic literature review. Thus, the need for systematic review has emerged. A comprehensive and systematic understanding of a certain area is very important to expand the knowledge of a particular topic and gives an idea of new research directions [10]. This study will help researchers to design new research topics and quickly get used to them. It also helps on identifying new research problems and topics that has not been investigated by analyzing the overall mode of the research that has been done to date. It will also give proper guidance for the researcher who wants to conduct studies on the topic of rural tourism. Indeed, [11] has answered this need by conducting a literature review using database in Web of Science published during 2009–2019. Their literature review is based on “sustainable rural tourism” key words. They reviewed 76 manuscripts with those key words. This article reviewed different papers related to factors affecting rural tourism from the Scopus database from 2001 to 2021 A.D. This study was conducted to answer the following questions: 1. What are the general characteristics related to rural tourism? 2. What is the pattern of use of selected articles in another research works? 3. What are the key factors affecting rural tourism?

2 Research Methodology 2.1 Search and Information Sources This study was conducted by reviewing the articles related to factors affecting rural tourism and homestay searched on the Scopus database. This study consists of the articles conducted from 2001 to 2021 A.D. to find a trend of factors affecting rural tourism and homestay and to provide the most important information. Different keywords were used to collect the articles related to the study. Keywords search process was: TIT − ABS − KEY((Factor ∗ OR Determinant∗) AND (Rural tourism OR homestay)) AND (LIMIT - TO(DOCTYPE, ar)) AND (LIMIT - TO(LANGUAGE, English)) AND (LIMIT - TO(SRCTYPE, j)) From above-mentioned key word 440 articles were found on the date 10th August 2022 in Scopus database. A detailed list of articles was extracted for further analysis.

A Systematic Literature Review on Factors Affecting Rural Tourism

47

2.2 Process of Data Collection and Article Screening In this study, only peer-reviewed journal articles were considered. Books, papers, conference papers, book chapters, and other editorial materials were excluded to make the study more systematic and scientific. Papers published in the English language only were considered for systematic review. The forms for data extraction have been developed to identify articles and provide an overview of research on rural tourism. Each article was then measured in a structured data extraction format by considering the quality of the study. From a total of 440 articles, first-level screening was done by dividing them into two parts, fundamentally related to factors affecting rural tourism or homestay and not related to rural tourism and homestay (Fig. 1).

Fig. 1. PRISMA framework

The result shows that 181 articles were relevant to the above-mentioned criteria. In the next step, the screening of data based on title and abstract was done. Finally, 61 articles covering the desired study as the main issue were selected for systematic literature review.

3 Results 3.1 General Characteristics To address question 1, analysis was done regarding the overall characteristics of selected studies. The analysis covered is a year-wise study in the related subject, journal-wise publication, methods applied, research perspective, country-wise study, citation by documents, citation by journals, citation by country, and co-occurrence of keywords. Year Wise Distribution The paper related to rural tourism has been steadily increased from 2005 to 2019 but there was a slight decrease in 2020 that may be due to Corona Pandemic (Fig. 2). In the

48

P. Sapkota et al.

Yearwise Distribuon 10

9

8

9

8 5

6 3

4 2 0

1

1

1

1

1

2

3

3

4

3

4

2006200720082009201020112012201320142015201620172018201920202021 Fig. 2. Year-wise distribution of articles

year 2021 we can observe gradual increment in the number of papers. It can be observed that the interest in rural tourism has been increasing tremendously over time. Journal-Wise Articles The systematic literature review was conducted by taking 61 articles published in different journals around the globe All the papers are selected from the Scopus database. Articles are taken from 47 different journals. Sustainability Journal (5 papers) and Tourism Management (3 papers) lead in the field of rural tourism, which is followed by Asian Academy of Management Journal, Advanced Science Letters, African Journal of Hospitality, Tourism and Leisure, Asia Pacific Journal of Tourism Research, DETUROPE, Journal of Travel and Tourism Marketing, Journal of Travel Research, Tourism Management Perspectives with 2-2 papers from each journal. The remaining journals had one paper each. Research Method and Tools Applied In the study of 61 different articles related to rural tourism, it was found that three different methods of research were applied (Table 1). The first method was quantitative analysis which was used in 44 articles (72.13% of the total study). Among these 44 articles, structural equation modeling (SEM) was the primary statistics method deployed to analysis the data collected. It was used in 16 articles. Followed by regression (including logistic regression) in the second. There were 9 documents which used regression in analyzing. Following descriptive statistics with or without correlation in 6 documents, analysis of variance (ANOVA) in 5, confirmatory factor analysis (CFA) in four articles, Exploratory Factor Analysis (EFA) in 3, canonical correlation in 2 articles, and importance-performance analysis (IPA) also in 2 articles. There were three documents that combine two or more analysis tools. Two documents used mixed method that combine EFA, CFA, and SEM. One article used three steps analysis, consists of principal correspondence analysis, cluster analysis, and discriminant analysis. The papers related to the quantitative analysis deployed questionnaire as data collection instrument. The second method was qualitative analysis which was used in 12

A Systematic Literature Review on Factors Affecting Rural Tourism

49

articles. Among the paper where qualitative analysis was applied, 9 papers use in-depth interview in collecting the data. Each of three remaining papers deploy focus group discussion, participants observation and structured qualitative questionnaire. The third was the mixed method which was applied in 5 papers. Table 1. Research method applied Classification Research methods

No.of articles

%

44

72.13

9

14.75

Quantitative analysis Survey questionnaire Qualitative analysis In-depth interview Focus group discussion

1

1.64

Participants observation

1

1.64

Qualitative questionnaire

1

1.64

5

8.19

61

100

Mixed analysis Mixed methods research design Total

Research Perspective Regarding the perspective of research out of 61 selected papers, it was found that 44 articles (equal to 72.13%) were written from a tourist perspective, 8 articles of them were from the stakeholder’s perspective, 5 papers from the resident’s perspective, and the remaining 4 papers from host’s perspective (Table 2). It shows that most of the articles were written from tourists’ perspective. Table 2. A research perspective

Research perspective

Classification

No. of articles

%

Tourists

44

72.13

Stakeholders

8

13.11

Residents

5

8.20

Hosts

4

6.55

61

100

Grand total

Region-Wise Classification It was found that the study related to rural tourism was conducted in various parts of the world. Most of the studies were done in Asia (50.81%) and Europe (40.98%) as shown in

50

P. Sapkota et al.

Table 3. The remaining studies were done in Africa and North America. It was observed that most of the studies were performed in Malaysia (11 papers) and it was followed by Spain (8 papers) and China (7 papers). Table 3. Region-wise classification Region

Number

%

Countries

Asia

31

50.81

Malaysia (11), China (7), South Korea (1), Taiwan (2), Indonesia (2), Thailand (3), Cambodia (1), India (2), Iran (1), Vietnam (1)

Europe

25

40.98

Spain (8), Portugal (4), Serbia (4), Cyprus (3), Czech Republic (1), Germany (1), Hungary (1), Turkey (1) Romania (1), UK (1)

Africa

3

4.92

Ghana (1), Kenya (1), South Africa (1)

North America

2

3.28

USA (2)

3.2 Citation and Keyword Co-occurrence Analysis To provide insight on research question 2, the study was applied to investigate documentwise, journal-wise, and country-wise citations from the selected 61 papers. Document-Wise Citations The number of citations in the papers represents the quality of documents, its influence and popularity within a research field. Out of 61 articles, one article entitled “Influence of the user’s psychological factors on the online purchase intention in rural tourism: Integrating innovativeness to the UTAUT framework” was found to have the highest number of citation (cited 340 times). Following in the second order is the article entitled “Factors for success in rural tourism development” (cited 326 times). The third place is the article entitled “The role of motivation in visitor satisfaction: Empirical evidence in rural tourism” with total citation 282 times. Articles which have citation more than 50 times are taken into consideration as shown in Table 4. Country-Wise Citations The study was conducted to find out country-wise citation as well. Countries with more than 10 citations for the selected study were considered for the study. It was observed that that the highest number was in Spain with 1140 citations from nine papers. This is followed by the USA with 438 citations (4 papers), Portugal 343 citations (6 papers), Malaysia 296 citations (13 papers), Taiwan 70 citations (2 papers), UK 58 citations (2 papers), Singapore 49 citations (1 paper), Cyprus 43 citations (3 papers), China 30

A Systematic Literature Review on Factors Affecting Rural Tourism

51

Table 4. Document-wise citations S. No.

Documents

Citations

1

san martín h. (2012)

340

2

wilson s. (2001)

326

3

devesa m. (2010)

282

4

loureiro s.m.c. (2008)

181

5

albacete-sáez c.a. (2007)

128

6

rasoolimanesh s.m. (2017)

107

7

jamal s.a. (2011)

105

8

murray a. (2015)

96

9

chen l.c. (2013)

69

10

kastenholz e. (2005)

69

11

loureiro s.m.c. (2010)

59

12

albaladejo-pina i.p. (2009)

58

13

greaves n. (2010)

52

citations (6 papers), Germany 22 citations (1 paper), Italy 22 citations (1 paper), Ghana 15 citations (1 paper), Czech Republic 14 citations (1 paper), Iraq 12 citations (1 paper), Romania 12 citations (1 paper), Russia 10 citations (3 papers), and Serbia 10 citations (3 papers). Keyword Co-occurrence Use of keywords which were related to rural tourism and homestay was also a concern. The keyword co-occurrence study always investigates the use of the frequent used keywords in the articles. This analysis was conducted with the help of co-occurrence. VOS viewer software was used to analyze the co-occurrence of keywords. The objective of the keyword co-occurrence study was to visualize trends of the important research topics in this field. We concentrated on the author keywords that appear below the abstract. This technique is used to count the papers in which two keywords comes together. In Fig. 3 the main keywords and their connection with other keywords (when node and keyword seem to be larger than results repetition of the keyword in more papers) can be observed. The line drawn between the nodes depicts the frequent co-occurrence of keywords in various papers, and the distance between nodes determines the relationship between keywords (shorter the distance, stronger the relation). In this study of 61 articles, considering the threshold of a minimum of ten occurrences, fifty-four keywords were found in eight main clusters. Each cluster is separated by different colors. The most frequently used keyword used in the study leading the cluster was “rural tourism” which represents the brown color. Other main keyword leading the clusters were: "satisfaction” (green), “rural development” (light blue), “tourism development” and “perception” (red),

52

P. Sapkota et al.

Fig. 3. Keyword co-occurrence map.

“hospitality industry” (yellow), “tourism management” (blue), “service quality” (grey), and “motivation” (purple). 3.3 Key Factors Examining 61 articles we noticed 53 different factors which affect rural tourism. The accommodation was the most important in rural tourism which was pointed 15 times in this study. Destination image (13) and price (13) were the two other most important factors that affect rural tourism (Table 5). Factors having more than seven frequencies are only considered for the study. Table 5. Key factors affecting rural tourism. S. No.

Factors

Frequency

1

Accommodation

15

2

Destination image

13

3

Price

13

4

Cultural experience

12

5

Security

11

6

Community participation

10

7

Adventure

9

8

Marketing

9

9

Peaceful environment

9

10

Leadership

8

11

Interaction

7

A Systematic Literature Review on Factors Affecting Rural Tourism

53

4 Discussion The aim of this paper was to examine the different studies related to factors affecting rural tourism. So, this paper examined the issues theoretically by observing the role of community, hosts, tourists, and stakeholders. This study analyzed the topic of factors affecting rural tourism, observing the pace in the literature as well as, multidimensional area that focus on the views regarding hosts, community, tourists, and other stakeholders. This research also focuses on the search for prior analysis of rural tourism structure with an emphasis on earlier studies. It was observed that systematic review was done on various other aspects of tourism but not on this topic. So, due to the lack of reviewed pieces of literature and relevance of systematic literature review and bibliometric approach this study was conducted. The outcomes of this study investigated the need for an integrated, environmental, social, and economic aspects of rural tourism from an interdisciplinary approach. The study reveals that after going through different papers, the questions could be prepared by giving priority on other and trends dimensions. It is also recommended to focus on different methodological instruments, and more applicable and empirical study to analyze more specific problems related to the topic. This study analyzed 61 articles from different journals published from January 2001– 2021 through a systematic process. It was observed that the trend of studies related to rural tourism is increasing significantly since the last decade. However, in comparison with other topics the literature is still scarce in this field. This paper is new in the framework and trends in the study of factors affecting rural tourism. Examination of the sources indicates that related studies have been done in various places of the world. However, most of the studies were conducted in Asia (50.81%) and Europe (40.98%). While going through the counties we found most studies were conducted in Malaysia (Asia). This proves that the study of rural tourism is considered as one of the major topics in the Asian region. The studies were conducted using qualitative, quantitative, and mixed methods of research where quantitative analysis was applied for most of the studies (72.13%). This shows most of the studies were conducted by researchers by preparing structured questionnaire and filling out it by respondents. While analyzing the perspective of research it was found that studies were conducted from tourists, hosts, residents, and stakeholder’s perspectives. Most of the studies (72.13%) were found to be done from a tourist perspective. This result emphasized the need for study from a tourist perspective for the success of rural tourism. The authors conducted different citation studies for the selected papers. From the study it was found that “Tourism Management” from which 3 papers are included got the highest number of citations (680). The other two journals got more than 200 citations "Journal of travel research” and “Journal of travel and tourism marketing”. The increasing interest in rural tourism can be observed in the number of citations of papers. Several factors have been identified as being critical to the success of rural tourism. Accommodation, destination image, and price are the key factors affecting rural tourism. So, concerned stakeholders need to focus on these key elements to achieve success in rural tourism. It was also observed that the keywords related to the study were divided into 8 clusters. Rural tourism was the main keyword related to the study. Whereas tourism,

54

P. Sapkota et al.

satisfaction, service quality, perception, motivation, and hospitality industry were other keywords leading each cluster.

5 Conclusion This study analyzed 61 articles from 47 journals published from January 2001–2021, related to factors affecting rural tourism through the data screening process. The study in rural documents shows that this area is quite wide and diversified so that a broader perspective needs to be included. This paper has indicated different dimensions of rural tourism and interesting trends in literature. The result that was obtained will help policy makers and stakeholders for the scientific development of rural tourism. Meanwhile, researchers can get some diverse topics that can open new research areas. The study has given more emphasis on the factors affecting rural tourism. It is essential to study rural tourism from a diverse perspective such as environmental, sociocultural, economic, etc. Some of the perspective related to factors affecting rural tourism has been studied but it would be better to include the impact of social media and technology factor as well in the future. This study has some limitations though it contributes a lot in rural tourism field. Firstly, the articles used in this research were only collected from the Scopus database. Thus, studies from other journals and publications which are not indexed in Scopus database on the same topic were not considered for this study. Only 61 out of available 440 articles were selected for the final study through the data screening process, which may also affect the result of the study. So, this may bound the opportunity to gain broader knowledge and information related to rural tourism. It is recommended that researchers monitor the development of the various keywords in literature and examine deeply some of the study clusters revealed in our research. Newspapers, books, and other informational sources could also be considered for new methodologies outside to organize and study various literatures. Future researchers should make a strong effort to generate greater knowledge and develop a new framework by examining the consensus as well as the limitations of the result of this study by doing a more scientific and systematic review in rural tourism.

References 1. Garrod, B., Wornell, R., Youell, R.: Re-conceptualising rural resources as countryside capital: The case of rural tourism. J. Rural Stud. 22(1), 117–128 (2006). https://doi.org/10.1016/j.jru rstud.2005.08.001 2. Tirado Ballesteros, J.G., Hernández Hernández, M.: Challenges facing rural tourism management: a supply-based perspective in Castilla-La Mancha (Spain). Tour. Hosp. Res. 21(2), 216–228 (2021). https://doi.org/10.1177/1467358420970611 3. Viljoen, J., Tlabela, K.: Rural tourism development in South Africa, trends and challenges. Hum. Sci. Res. Counc. 1–29 (2007) 4. Nair, V., Munikrishnan, U.T., Rajaratnam, S.D., King, N.: Redefining rural tourism in malaysia: a conceptual perspective. Asia Pacific J. Tour. Res. 20(3), 314–337 (2015). https:// doi.org/10.1080/10941665.2014.889026

A Systematic Literature Review on Factors Affecting Rural Tourism

55

5. Wilson, S., Fesenmaier, D.R., Fesenmaier, J., Van Es, J.C.: Factors for success in rural tourism development. J. Travel Res. 40(2), 132–138 (2001). https://doi.org/10.1177/004728750104 000203 6. Devesa, M., Laguna, M., Palacios, A.: The role of motivation in visitor satisfaction: empirical evidence inrural tourism. Tour. Manag. 31(4), 547–552 (2010) 7. Látková, P., Vogt, C.A.: Residents’ attitudes toward existing and future tourism development in rural communities. J. Travel Res. 51(1), 50–67 (2012). https://doi.org/10.1177/004728751 0394193 8. Sharpley, R.: Host perceptions of tourism: a review of the research. Tour. Manag. 42, 37–49 (2014) 9. Nunkoo, R., Ramkissoon, H.: Travelers’ E-purchase intent of tourism products and services. J. Hosp. Mark. Manag. 22(5), 505–529 (2013). https://doi.org/10.1080/19368623.2012.680240 10. Hulland, J., Houston, M.B.: Why systematic review papers and meta-analyses matter: an introduction to the special issue on generalizations in marketing. J. Acad. Mark. Sci. 48(3), 351–359 (2020). https://doi.org/10.1007/s11747-020-00721-7 11. An, W., Alarcón, S.: How can rural tourism be sustainable? a systematic review. Sustain. 12(18) (2020). https://doi.org/10.3390/SU12187758

K-Modes with Binary Logistic Regression: An Application in Marketing Research Jonathan Rebolledo1 and Roman Rodriguez-Aguilar2(B) 1 Facultad de Ingeniería, Universidad Anáhuac México, Av. Universidad Anahuac 46, Lomas

Anahuac, 52786 Lomas Anahuac, Mexico 2 Facultad de Ciencias Económicas y Empresariales, Universidad Panamericana, Augusto

Rodin 498, 03920 Mexico City, Mexico [email protected]

Abstract. Binary logistic regression is a statistical method used to analyze data with binary outcome variables. It is a type of generalized linear model (GLM) and is often used in fields such as medicine, psychology, and sociology. Using this model with some degree of data noise might give wrong results. Keywords: k-modes · Outliers · Logistic regression · Deviance

1 Introduction Binary logistic regression is a statistical model commonly used in various fields, including social sciences, medical research, and marketing, to analyze and understand the relationships between variables. The model uses a logistic function to map the linear combination of predictor variables to a probability value between 0 and 1. The model estimates the parameters of the logistic function using maximum likelihood estimation, which finds the values of the parameters that maximize the likelihood of the observed data. The model output includes coefficients for each predictor variable, which represent the strength and direction of the relationship between the variable and the outcome. Binary logistic regression is a powerful tool for analyzing and predicting binary outcomes, but it is important to consider its assumptions and limitations, such as linearity of the relationship between predictor variables and the outcome and independence of observations. 1.1 Outliers in Binary Logistic Regression Outliers can affect binary logistic regression in the same way they affect linear regression. Outliers are data points that are far away from the bulk of the data and can have a disproportionate effect on the model. Outliers can impact the estimated coefficients and increase the variability of the model, leading to biased and unreliable predictions. In binary logistic regression, outliers may lead to an incorrect classification of the outcome variable for some observations, which can lead to inaccurate predictions. For © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 56–64, 2024. https://doi.org/10.1007/978-3-031-50158-6_6

K-Modes with Binary Logistic Regression: An Application

57

example, an outlier with a very high probability of the outcome variable may shift the classification threshold and result in misclassifying other observations. Identifying outliers in categorical data can be challenging, as there are no clear numerical cut-off points. However, there are several methods that can be used: Frequency distributions: One way to identify outliers in categorical data is to examine the frequency distributions of each category. If a category has a much lower or higher frequency than the others, it may be an outlier. Box plots: Another way to identify outliers in categorical data is to create box plots for each category. Box plots show the median, quartiles, and range of the data, and can help to identify categories that have significantly different distributions from the others. Chi-square test: A chi-square test can be used to compare the observed frequencies of each category to the expected frequencies, based on the overall distribution of the data. Categories that have a significantly higher or lower observed frequency than expected may be outliers. Once outliers have been identified, there are several ways to deal with them: • Removing outliers: This is the most common method of dealing with outliers. However, it should be done with caution, as removing too many observations may lead to a loss of information. • Transforming the data: In some cases, outliers can be caused by a skewed distribution. Transforming the data, such as taking the log or square root, can help to alleviate this problem. • Using robust estimation methods: Some estimation methods, such as M-estimators and maximum likelihood estimators, are less sensitive to outliers than traditional methods. This includes the Robust Regression: a type of regression technique that uses weighted least squares to minimize the effect of outliers in the data. It works by giving higher weights to data points that are close to the line of best fit and lower weights to data points that are further away. • Regularization involves adding a penalty to the cost function to reduce the effects of outliers on the model. This penalty is determined by how far away each data point is from the line of best fit. Some examples of regularization techniques include Ridge Regression, which adds an L2 penalty to the cost function, and Lasso Regression, which adds an L1 penalty to the cost function. 1.2 Categorical Variables In binary logistic regression, categorical variables can be used as predictor variables to explain the variation in the outcome variable, which is binary (0 or 1). There are two types of categorical variables in binary logistic regression: Nominal variables: These are variables that have no inherent order. For example, a variable that represents different types of fruits (apple, banana, orange) is a nominal variable. In binary logistic regression, nominal variables are often represented by dummy variables. A dummy variable is a binary variable that represents the presence or absence of a specific category. Ordinal variables: These are variables that have an inherent order. For example, a variable that represents different levels of education (high school, college, graduate) is an ordinal variable. In binary logistic regression, ordinal variables are often represented by

58

J. Rebolledo and R. Aguilar

ordinal dummy variables. An ordinal dummy variable is a binary variable that represents whether an observation is in a specific category.

2 Marketing Research In marketing research, almost all the data is categorical. Why? Because in real life, an individual buys or not a product or service based on the properties of the product or service, they like or not the price, they like or not the design, they like or not the color, etc. So when evaluating a marketing campaign, 95% or more of the predictive variables are categorical, which includes demographical variables such as gender, socioeconomical level, age in range levels, the location of the individual, etc. And as mentioned before, the Binary Logistic Regression model may give them incorrect results since it has not been optimized for working with all predictive variables being categorical. 2.1 K-Modes and Binary Logistic Regression One of the assumptions in the basic model is that there is one big population and there are no outliers, but actually a lot of outliers are present in the data set. How can we tell to the basic model that there are many outliers? How can we obtain robust estimations of the Beta parameters? We want to know the ranking and the Odds of the principal variables, without removing the “outliers” since every registry in the data set has money implications, the client has paid a lot of cash for the data, and removing an individual is not an option. The clients use the Odds ratio estimation because of the interpretation in terms of percentage, it is easy to understand that the possibilities of purchasing a product or service increases “X percentage” if an individual likes a specific characteristic of the product or service they are selling. There are a lot of methods to detect patterns in statistics, these are known as “unsupervised learning”; these methods allow them to detect such patterns. The basic idea is to add an additional variable to the dataset indicating the group each individual belongs to, then run the binary logistic regression with this new variable and obtain the best estimation of the other predictive variables in order to determine the path the client should follow to increase the probability of the dependent variable. An obvious question arises: how many groups do we need? This is obtained by analyzing the pattern of the deviance as a goodness of fit of the model; in this approach (since all their variables in the dataset are categorical dummy), the use of K-Modes unsupervised learning is applied, with the number of groups varying from 2 to 10. K-modes is a clustering algorithm used for categorical data. It is similar to the kmeans algorithm, but instead of calculating the mean of the data points in a cluster, it calculates the mode. The mode is the most frequently occurring value in a dataset. The algorithm starts by randomly selecting k initial modes, one for each cluster, and then assigns each data point to the cluster with the closest mode. The modes are then recalculated for each cluster, and the process is repeated until the assignments of data points to clusters no longer change. k-modes is used for categorical data because it can handle categorical variables that have many different levels.

K-Modes with Binary Logistic Regression: An Application

59

To determine the optimal number of groups, we create a scatter plot of the behavior of the deviance for 1 group (the basic model), 2 groups, 3 groups, etc. then we select the number of groups where the deviance is minimized. Later we run the binary logistic regression with this additional variable and compare the results. How do we know this methodology actually works? Since there are many approaches, we compare the basic logistic regression model, the ridge regression, and this new approach with K-modes; we cannot use “stepwise” o Lasso regularization because in marketing research the variables evaluated by the individual cannot value cero, because of the money implications; These variables were obtained by qualitative studies, which indicate these attributes must be evaluated. Sad but true. Ultimately, we use the Lasso regularization as validation of the path of the top 4 variables the client must follow to increase the sales of the product or service.

3 Data Set The data set belongs to IPSOS©, it corresponds to a study of evaluating the best fast-food restaurant with the main question “What makes an individual choose between one place to another?”. The data base consists of 4,512 interviewees. The survey was carried out in a period of 1 month (March 2022) in the 3 main Metropolitan Areas of Mexico: Mexico City, Guadalajara and Monterrey. Men and women from 18 to 45 years of all socioeconomic levels were considered. Each interviewee evaluates the brands they know and have consumed, being able to evaluate only 1 brand. The filter is that none of the interviewees works in any market research company, as well as having consumed products from any of the following brands in the last 20 days: • • • • • • • • • •

Benedetti’s Pizza Burger King Domino´s Pizza KFC Little Caesar’s McDonald’s Papa John’s Pizza Hut Subway Carl’s Jr

Dependent variable is purchase consideration: Would you consider buying again in “xxx” soon? Possible responses “yes”, “no”. The base structure and some basic demographical insigns can be seen in Tables 1 and 2.

4 Results After executing the K-modes from 1 to 10 groups, we see this pattern graphically:

60

J. Rebolledo and R. Aguilar Table 1. Data base structure.

Variable Respondent_Serial Serial number resp_gender Are you male of female?

Label

Measurment Scale Nominal Nominal

resp_age

Respondent Age

Ordinal

Brand Q8 Q11_1 Q11_2 Q11_3 Q11_4 Q11_5 Q11_6 Q11_7 Q11_8 Q11_9 Q11_10 Q11_11 Q11_12

Evaluated Brand Purchase consideraon It has the best promoons It offers a good variety of items on the menu It has a fast service It features quality food\ingredients It has quality service (friendly\ courteous staff) It has the tasest food It gives me more value for my money It's a fast food place I like Their poron sizes\dishes are large enough to sasfy my hunger It's a clean place to visit It's easy to order and get my food It's a restaurant for people like me

Nominal Target, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary Input, Binary

Possible values Numerical 1 Male, 2 Female 1 (18-24 years), 2(25-34 years), 3(35-45 years) Text 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes 0 No, 1 Yes

Table 2. Demographics statistics of the data set.

Male Female Total

Gender Frequency Percent 2254 49.956 2258 50.044 4512 100.0

1 Mexico City 2 Guadalajara 3 Monterrey Total

Location Frequency Percent 2256 50.0 1128 25.0 1128 25.0 4512 100.0

1.00 18-24 years 2.00 25-34 years 3.00 35-45 years Total

Age Frequency Percent 1940 43.0 1710 37.9 862 19.1 4512 100.0

Evaluated Brand Frequency 1 Benedetti's Pizza 278 2 Burger King 517 3 Carl’s Jr 466 4 Domino´s Pizza 501 5 KFC 539 6 Little Caesar’s 555 7 McDonald’s 457 8 Papa John’s 299 9 Pizza Hut 427 10 Subway 473 Total 4512

Percent 6.161348 11.45833 10.32801 11.10372 11.94592 12.30053 10.12855 6.626773 9.463652 10.48316 100

The number of groups that minimize the deviance is 5 with a deviance value of 5092.949 (Fig. 1), then the logistic regression with k = 5 is conducted.

K-Modes with Binary Logistic Regression: An Application

61

Fig. 1. Pattern of the deviance after including the K-Modes variable

We compare the Beta estimations, significance, and Odds ratio (>1) of 4 models: K-Modes with Logistic Regression, Basic Binary Logistic Regression, Ridge regularization and additionally the LASSO regularization, this one not useful due to money implications. 4.1 K-modes with Binary Logistic Regression See Table 3. 4.2 Basic Binary Logistic Regression See Table 4. 4.3 Ridge Regression Deviance: 5316.5, Optimal lambda 0.027543 (# of folds for CV: 10, seed 1234) (Table 5). 4.4 LASSO Deviance: 5215.25, Optimal lambda 0.003814 (#of folds for CV; 10, seed 1234) (Table 6). As we can see, the K-modes has the best goodness of fit, the deviance is much lower than the others, the beta estimations are better and the odds of the variables. One thing to highlight is the significance (α = 0.05) of the betas, in the Basic Binary Logistic Regression only Q11_1 is statistically important and that is weird since this set of variables were defined by previous qualitative studies.

62

J. Rebolledo and R. Aguilar Table 3. K-Modes optimization results.

Variable (Intercept) Q11_1Yes Q11_2Yes Q11_3Yes Q11_4Yes Q11_5Yes Q11_6Yes Q11_7Yes Q11_8Yes Q11_9Yes Q11_10Yes Q11_11Yes Q11_12Yes Cluster 1 Cluster 2 Cluster 3 Cluster 4

Esmate 0.402064 0.418671 -1.2806 0.167329 0.266051 -0.9973 0.231491 0.219134 0.171757 0.014773 -1.20368 0.288735 -1.23185 -1.95935 2.899163 -1.08823 -1.39291

Pr(>|z|) 0.137195 1.34E-05 4.6E-23 0.095285 0.006775 2.55E-13 0.025803 0.025642 0.108669 0.881213 1.31E-18 3.97E-03 1.02E-20 1.05E-10 1.36E-21 0.000126 1.21E-07

Coefficients: Odds Rao Variable Importance *** *** *** *** *** ***

*** *** *** *** *** *** ***

52% 18% 30% 26% 24% 19% 1% 33%

13% 2% 10% 12% 3% 11% 11% 11% 9% 3% 12% 3%

Ranking 1 12 7 3 9 4 5 6 8 10 2 11

Table 4. Binary logistic regression results.

Variable (Intercept) Q11_1Yes Q11_2Yes Q11_3Yes Q11_4Yes Q11_5Yes Q11_6Yes Q11_7Yes Q11_8Yes Q11_9Yes Q11_10Yes Q11_11Yes Q11_12Yes

Esmate -1.36636 0.376158 -0.00903 0.130001 0.040665 0.015428 0.019338 0.139824 0.129904 -0.15639 0.058095 0.135644 -0.01225

Pr(>|z|) 2E-16 *** 5.79E-06 *** 0.9044 0.1105 0.6139 8.53E-01 0.8205 0.0979 0.1381 0.0556 0.4555 1.04E-01 0.8756

Coefficients: Odds Rao Variable Importance 46% 14% 4% 2% 2% 15% 14% 6% 15%

11% 8% 9% 8% 8% 8% 9% 9% 7% 8% 9% 8%

Ranking 1 10 4 7 9 8 2 5 12 6 3 11

In this study, four different statistical methods were used to analyze the data, Kmodes analysis, basic logistic regression, ridge regularization, and LASSO regularization. The results of each method reveal different variable importance rankings, which are as follows:

K-Modes with Binary Logistic Regression: An Application

63

Table 5. Ridge regularization results.

Variable (Intercept) Q11_1Yes Q11_2Yes Q11_3Yes Q11_4Yes Q11_5Yes Q11_6Yes Q11_7Yes Q11_8Yes Q11_9Yes Q11_10Yes Q11_11Yes Q11_12Yes

Coefficients: Esmate Odds Rao Variable Importance -0.52174 -0.22532 25% 10% -0.01045 1% 8% -0.10524 11% 9% -0.03615 4% 8% -0.0332 3% 8% -0.03678 4% 8% -0.10317 11% 9% -0.09653 10% 9% 0.052596 7% -0.04786 5% 8% -0.10396 11% 9% -0.01914 2% 8%

Ranking 1 11 2 8 9 7 4 5 12 6 3 10

Table 6. LASSO regularization results.

Variable (Intercept) Q11_1Yes Q11_2Yes Q11_3Yes Q11_4Yes Q11_5Yes Q11_6Yes Q11_7Yes Q11_8Yes Q11_9Yes Q11_10Yes Q11_11Yes Q11_12Yes

Coefficients: Esmate Odds Rao Variable Importance -0.55655 -0.32161 38% 20% 0 -0.09726 10% 16% 0 0 0 -0.08244 9% 16% -0.09433 10% 16% 0 -0.0155 2% 15% -0.10669 11% 16% 0

Ranking 1 3

5 4 6 2

K-modes analysis: The most important variables in order are Q11_1, Q11_11, Q11_4, and Q11_6. Basic logistic regression: The most important variables in order are Q11_1, Q11_7, Q11_11, and Q11_3. Ridge regularization: The most important variables in order are Q11_1, Q11_3, Q11_11, and Q11_7. LASSO regularization: The most important variables in order are Q11_1, Q11_11, Q11_3, and Q11_8.

64

J. Rebolledo and R. Aguilar

While the variable Q11_1 is consistently ranked as the most important variable across all four methods, the other variables’ rankings differ. These findings provide valuable insight into which variables are more influential depending on the statistical method used, allowing for more informed decision-making and further analysis. K-modes is the winner; the Lasso (that cannot be used to communicate results to the client) also confirms the key variables (Q11_1 and Q11_11) that must be considered if the client wants to increase the probability of purchase consideration being “It has the best promotions” and “It’s easy to order and get my food” as the most important variables with a variable relative importance of 13% and 12% respectably, the odds are also very High.

5 Conclusions The methodology here presented is one alternative when the data set has categorical variables and a huge number of outliers that can mislead the conclusions of the model. This alternative represents something easy to use, easy to implement and computationally affordable. In case the database has mixed types of variables, (categorical and numerical) then the k-Prototypes Clustering could be used, and the minimization occurs, in case of scale variable target, using RMSE as goodness of fit of the model (suggested).

References Agresti, A.: Categorical data analysis (3rd ed.). John Wiley & Sons (2013) Bishop, C.M. (ed.): Pattern Recognition and Machine Learning. ISS, Springer, New York (2006). https://doi.org/10.1007/978-0-387-45528-0 DeSarbo, W.S., Ramaswamy, V., Cohen, S.H.: Market segmentation with choice-based conjoint analysis. Mark. Lett. 6(2), 137–147 (1995) Everitt, B.S., Hothorn, T.: An introduction to applied multivariate analysis with R. Springer (2011) Field, A.: Discovering statistics using IBM SPSS statistics (5th ed.). SAGE Publications (2018) Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E.: Multivariate data analysis (8th ed.). Cengage Learning (2018) Hastie, T., Tibshirani, R., Friedman, J.: The elements of statistical learning: Data mining, inference, and prediction (2nd ed.). Springer (2009) Hosmer, D. W., Lemeshow, S., & Sturdivant, R.X.: Applied logistic regression (3rd ed.). John Wiley & Sons (2013) Huber, P.J.: Robust Statistics. John Wiley & Sons (1981) Huang, Z.: A fast-clustering algorithm to cluster very large categorical data sets in data mining. Res. Issues Data Min. Knowl. Discov. (DMKD ‘97) (1997) James, G., Witten, D., Hastie, T., Tibshirani, R.: An introduction to statistical learning: with applications in R. Springer (2013) Jolliffe, I.T., Cadima, J.: Principal component analysis: a review and recent developments. Philos. Trans. Royal Soc. A: Math., Phy. Eng. Sci. 374(2065), 20150202 (2016) Kelleher, J.D., Mac Namee, B., D’Arcy, A.: Fundamentals of Machine Learning for Predictive Data Analytics: Algorithms, Worked Examples, and Case Studies. MIT Press (2015) Malhotra, N.K.: Marketing research: an applied orientation (7th ed.). Pearson (2019) McCullagh, P., Nelder, J.A.: Generalized linear models (2nd ed.). Chapman and Hall/CRC (1989) Tibshirani, R.: Regression shrinkage and selection via the Lasso. J. Roy. Stat. Soc.: Ser. B (Methodol.) 58(1), 267–288 (1996)

Optimizing Headways Using Evolutionary Algorithms for Improved Light Rail Transit System Efficiency and Passenger Service Oomesh Gukhool(B)

, Nooswaibah Binti Nooroodeen Soosor, Ravindra Boojhawon, and Mohammad Zaheer Doomah University of Mauritius, Reduit, Mauritius [email protected]

Abstract. This study compares and evaluates the effectiveness of two optimization algorithms, Genetic Algorithm (GA) and Differential Evolution (DE), in determining the optimal headway values for Light Rail Vehicles (LRVs) at stations. The aim is to minimize both passenger waiting time and operational costs for LRT systems. Both algorithms use a standard cost function, where the decision variables are headway values representing the time intervals between the arrival of consecutive LRVs at the same station. The study assesses the performance of both algorithms and determines which is more effective in optimizing headways for LRVs. The results of this study will provide valuable insights to LRT operators in terms of choosing the most efficient algorithm to optimize headways, which can lead to reduced operational costs and improved service quality for passengers. Ultimately, optimizing LRV headways can help LRT systems to enhance their efficiency and provide better service to their passengers. Keywords: Headway · Optimization · Evolutionary Algorithms

1 Introduction Headway optimization is a critical component of public transportation, as it pertains to the time interval between consecutive vehicles on a given route or line. This interval determines the frequency of service and the number of vehicles necessary to meet demand. Lately, there has been an increasing interest in leveraging data-driven techniques to optimize headways and enhance transit operations. One specific application of headway optimization is in light rail transit (LRT) systems. LRT systems are electric railway systems that typically operate on tracks in dedicated lanes or in mixed traffic and serve intermediate distances and denser urban areas. These systems frequently have multiple lines, each with its schedule and headway. LRT headway optimization involves utilizing real-time data and predictive analytics to adjust the frequency of service based on passenger demand, traffic conditions, and other relevant factors. This approach can reduce wait times for passengers, improve reliability, and increase overall system efficiency. Many research studies have delved © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 65–75, 2024. https://doi.org/10.1007/978-3-031-50158-6_7

66

O. Gukhool et al.

into the utilization of data-driven techniques for optimizing train headways, and some of these studies have employed machine learning algorithms to dynamically adjust the headways in response to changing conditions. [1, 2]. However, it has been observed that there is a paucity of similar studies focusing specifically on LRT systems. The application of data-driven approaches to optimize headways in LRT systems remains relatively unexplored, warranting further research in this area to bridge the gap and uncover potential opportunities for enhancing the efficiency of LRT operations. While there is no concrete research conducted in the field of LRT headway optimization, we have seen studies conducted in other transit system that supports the importance of optimizing headway to improve operations and passenger demand. Several studies have contributed to the field of transit headway optimization. For instance, one study investigated how stochastic vehicle arrivals impacted optimal stop spacing and headway for a feeder bus route [3]. Another study proposed a multi-objective model for scheduling problems in Bus Rapid Transit (BRT) systems [4]. Other studies have explored the influence of value-of-time distributions on transit headway optimization [5], the combined effects of probabilistic vehicle arrivals and passenger walking time on transfer coordination [6], and the optimization of headways with stop-skipping control in BRT systems [7]. Furthermore, research has been conducted on optimal control of headway variation on transit routes [8] and multi-objective optimal formulations for bus fleet size in public transit systems under headway-based holding control [9]. Despite the lack of application of optimization algorithms such as genetic algorithms (GA) and differential evolution (DE) to optimize LRT headway, there have been notable studies that have successfully utilized GA and DE techniques to optimize headways in other transit systems. For instance, a study presented in [4] introduced a novel multiobjective model for Bus Rapid Transit (BRT) scheduling problems, which was solved using the Fast Non-dominated Sorting Genetic Algorithm (NSGA-II), a commonly used multi-objective optimization algorithm. Another study conducted by [10] proposed a two-phase algorithm that combined a single-objective covariance matrix adaptation evolution strategy with a multi-directional local search to optimize headways in complex urban mass rapid transit systems. The authors of [11] introduced a novel bi-level optimization model for solving the transit frequency setting problem in bi-modal networks. The proposed model leverages a differential evolution (DE) algorithm in the upperlevel model to effectively determine the optimal headways for a given route structure, showcasing its innovative approach to tackling this complex problem. These studies collectively demonstrate the potential of optimization algorithms, such as GA and DE, in optimizing headways in transit systems, and highlight the need for further research in this area to develop more robust and efficient optimization techniques for LRT headway optimization. Despite this, there has been no direct comparison of the two algorithms for optimizing headway in LRT systems. DE has been shown to have faster convergence than GA for a wide range of problems, particularly those with highdimensional search spaces, due to its more directed exploration of the search space through the mutation operator [12, 13]. Furthermore, DE requires less parameter tuning compared to GA, which has several parameters that need careful calibration, including the selection operator, crossover operator, mutation operator, and population size. In contrast, DE only requires tuning of a few parameters, such as the mutation factor and

Optimizing Headways Using Evolutionary Algorithms

67

crossover rate [12, 13]. DE is also more robust than GA in handling changes in the cost function or search space due to its fixed mutation and crossover operator, while GA’s more complex operators may be sensitive to such changes [12, 13]. DE is also more effective in incorporating constraints directly into the cost function compared to GA, which requires a more complex approach to handling constraints [12, 13]. The aim of this study is to compare and evaluate the efficacy of two algorithms, GA and DE, in optimizing headways for Light Rail Vehicles (LRVs) with the goal of minimizing passenger waiting time and operational costs for LRT systems. Both algorithms utilize a standard cost function, in which the decision variables are headway values, representing the time intervals between consecutive LRV arrivals at the same station. By comparing the performance of these two algorithms, we can determine which algorithm is more effective at optimizing LRV headways in order to reduce LRT operation costs and passenger waiting times. By optimizing the headway values using a cost function and an optimization algorithm, LRT systems can improve their efficiency and reduce costs, while providing better service to passengers.

2 Methodology The proposed optimization algorithms are used to determine the best solution to a cost function which minimizes both the average waiting time of passengers and the service operator costs. The decision variables are the headways, and the optimization algorithms used are DE and GA, both implemented in MATLAB. The algorithms’ performance is assessed by analyzing their capability to generate high-quality solutions, their rate of convergence, and their robustness in dealing with constraints. Statistical analysis techniques, such as t-tests, are employed in the evaluation process. The results of the study shed light on the performance of both algorithms, which are elaborated in detail in the result and discussion section. 2.1 Cost Function The defined mathematical model considers the movement of LRV along only one direction of the LRT line. The LRV move to terminal station S starting from origin station 1. The daily operating time of the LRT service is truncated into P equal-length periods. In the model, stations are denoted by s, where s can take values between 1 and S-1 inclusive, and the periods are denoted by p, where p can take values between 1 and P inclusive. Moreover, in this case study, the length of each period is taken to be 1 h, and the start of each period p in the model occurs at multiples of 60 starting from 0th minute in period 1, 60th minute in period 2, and so on. The first LRV in each period depart from origin station 1 at the start of each period. The cost function Z incorporates two conflicting objectives: the passenger waiting time–cost function and the service operator cost function. The relative importance of each objective is determined by the weights W0 and W1 . W0 and W1 are user-defined values between 0 and 1; sum to 1; and are set at the discretion of the service operator and relevant stakeholders. The passenger waiting time model follows the assumptions that passengers arrive randomly at stations and that they board on the first arriving LRV at

68

O. Gukhool et al.

the respective stations they are waiting at. The mean waiting time per passenger waiting in period p is then assumed to be half the headway value in that period in accordance with the passenger model defined by Bowman and Turnquist [14]. The cost function Z is thus defined as follows: ⎛ ⎞ ⎞ ⎛  S−1 P  P   1 ωps × hp ⎠ + ⎝W1 × λ × np ⎠ (1) min Z = W0 × β⎝ 2 p=1 s=1

p=1

and in some of the tests below, is subject to the constraint:   60p − start of period p at station 1 < Tmax np = floor hp

(2)

The decision variables of the model are the headway values hp to be set in each period p to minimize both the passenger waiting times and the total service operator costs. hp lies in the interval [hlb … hub ] where hlb is the minimum value that hp can take so as to have a safe running distance between any two LRV and hub is an operator-decided value which represents the maximum passenger waiting time allowed during a period p. Moreover, between any two periods p and p + 1, the mathematical model ensures that there is at least hlb minutes between the last departing LRV in period p and the first departing LRV in period p + 1, by accordingly adjusting the beginning of period p + 1, if need be. ωps denotes the number of passengers waiting at station s in period p; β is the dollar equivalent per minute of passenger waiting time; np is the number of LRV serving the direction of the LRT line under study during period p; Tmax represents the maximum available number of LRV to serve the direction of the LRT line under study during any period of the day; λ denotes an average fixed cost of running of an LRV along the LRT line. It encapsulates the per-LRV operating expenses, salaries of the LRT staff, maintenance cost and so on. Two different optimization algorithms are considered to minimize Z, namely GA and DE. 2.2 Optimization Algorithms GA is a heuristic search method which mimics the processes involved in the biological evolution theory. It is based on the principle that only the fittest individuals survive from one generation to the next generation. During the initial iteration step (generation) of the algorithm, a set of parent individuals is produced by random generation. Each of these individuals represents a potential solution to the cost function. The population size, which is a value determined by the user, refers to the quantity of parent individuals generated at each iteration step. The parent individuals then undergo the crossing and mutation operations in an attempt to generate better offspring individuals, that is, better candidate solutions to the cost function. After each iteration step, the fitness of both parents and offspring individuals is evaluated to minimize the cost function. The fitter individuals are then chosen as parents for the next generation, and this process continues until a user-defined termination criterion is met. In this study, we set the termination criteria as a maximum number of iterations after which the algorithm stops.

Optimizing Headways Using Evolutionary Algorithms

69

DE is categorized an evolutionary algorithm [15]. DE follows a similar approach to GA by creating a population of potential solutions, with the population size at each iteration step determined a user-defined value. While it shares commonalities with GA in that both algorithms use genetic operators such as mutation, crossover and selection on possible candidate solutions, DE distinguishes itself from GA in the mechanism that it employs to do these operations. In this study, the classic DE/rand/1/bin strategy was employed as the DE approach. During the mutation stage, mutant vectors are created for each member of the population known target vectors. This is done by carrying out a weighted difference using three distinct random individuals (other than the target vector) from the current population. In the crossover stage, new individuals (trial vectors) are created by mixing components of mutant vectors with components of their corresponding target vectors according to some defined probabilistic rule. The fitter individual between the trial vector and its corresponding target vector is chosen as a member of the next population in each iteration. This process is repeated until the algorithm reaches the user-defined termination criterion. For the purpose of this study, we set the termination criterion as the maximum number of iterations after which the algorithm stops. For DE, the codes, as shown in Fig. 2, were written in reference to the steps outlined by [15], whereas the codes used for GA (Fig. 1) have been developed by [16]. For both algorithms, the codes were adjusted so that the generated candidate solutions at the first iteration were integer values from the interval [hlb … hub ]. Moreover, after the mutation stage for GA and the crossover stage for both GA and DE, the offspring candidates were bounded between hlb and hub and rounded off to the nearest integer values.

Fig. 1. Genetic algorithm

We examined the performance of the two algorithms under two different conditions:

70

O. Gukhool et al.

Fig. 2. Differential evolution algorithm

1. Without the constraint defined by Eq. (2) above. 2. With the defined constraint in Eq. (2) above. To this end, the constraint is added to the cost given by

function,  Eq. (1), through the use of a penalty function defined by M Pp=1 max 0, np − Tmax , where the penalty coefficient M is set to 10,000 and Tmax is 4. In both conditions 1 and 2, for different values of (population size/maximum number of iterations), the optimal values obtained by both algorithms were recorded over 50 runs of the algorithm. The values for (maximum number of iterations/population size) tested were: (50/100), (100/100), (100/150) and (150/150). The two-sample one tailed t-test (assuming inequality of variances) was then carried out on the values recorded for each combination of (maximum number of iterations/population size) considered, in order to determine if there was statistical evidence that one algorithm outperforms the other. The O-D matrix information was based off data obtained from the local LRT service provider, Metro Express Limited (MEL). The model parameter values were set as follows: W0 = 0.5, W1 = 0.5, λ = $10, P = 13, S = 19 and β = 0.0067. The GA parameters (gamma, mu, beta) as defined in the codes by Heris [16] regulate the crossover and mutation processes, and were set to (0.05, 0.08, 0.08). For DE, the scale factor F, and the crossover rate CR, were set to (F, CR) = (0.2, 0.05). The choice for the combination of the parameter values used for DE and GA has been made after the following investigations had been carried out: • For a particular cost function (other than the above-defined one), an extensive trial and error runs over different combinations of parameter values was carried out.

Optimizing Headways Using Evolutionary Algorithms

71

• The set of parameter values which yielded the minimum recorded cost function was then chosen as the final set of parameter values.

3 Results and Discussion 3.1 Quality of Solution To provide evidence of the effectiveness of the optimization model and both optimization algorithms in decreasing the overall costs of an LRT network, we conducted a survey of the local LRT network and collected data on headway values during each time period. The generalized cost of the LRT network under the current operational plan was calculated and compared with the optimal cost values obtained by both DE and GA (using the cost function with constraints at (population size/maximum number of iterations) = (150/150)). It was observed that both DE and GA were successful in decreasing the generalized cost of the LRT network by 6.09% and 6.15% respectively, indicating the efficacy of the model and both optimization algorithms at reducing the generalized cost of the LRT network. In contrast to our initial observation of a consistent headway for LRVs during our on-site research, we discovered that both optimization algorithms produced varying headway values for each time period based on the number of passengers waiting at each station. Moreover, the optimized headway resulted in a decrease in overall costs. Table 1 presents a summary of the optimal costs achieved through four tests that employed various population sizes and iterations in both the DE and GA optimization algorithms, using the cost function (Eq. (1)) as previously described. To examine whether the mean value of the best solution for GA is inferior to that of DE, taking into account the constraints in their respective cost functions, t-tests were conducted. A one-tailed test was conducted because the alternative hypothesis was directional. Four tests were carried out with varied iterations and population size. In all four tests, the degrees of freedom (df) were 49, and the one-tailed t-tests were performed. The t critical one-tail values were 1.675905025 in test 1, 1.676550893 in tests 2–4, and the alpha level was set at 0.05. The computed t-statistic exceeded the critical t-value at a significance level of 0.05, indicating that we can reject the null hypothesis and accept the alternative hypothesis. This indicates that the mean value of the best solution for GA was considerably lower than that of DE, implying that GA outperformed DE in discovering the optimal solution. Moreover, the p-value for each test was very small (less than 0.05), indicating that the observed differences in the means were statistically significant. Thus, we can conclude that there is strong evidence that GA outperformed DE in finding the optimal solution, and this holds true for different combinations of maximum iteration and population size when considering the cost function with constraints. 3.2 Convergence Speed One common way to assess convergence speed is to plot the values of the cost function To study the convergence rate of each algorithm, a plot of the minimum cost recorded per iteration was made for both algorithms. Figs. 3 and 4 show graphs of the best minimum cost recorded at each iteration for a random run of GA and DE at (population size/maximum number of iterations) = (150/150). In this case, the cost function was

72

O. Gukhool et al. Table 1. Best average cost for DE and GA (with constraint)

Test

Max iteration

1

100

2

100

3

150

4

150

Population size

Best average cost DE

Best average cost GA

50

659.31

652.11

100

653.3

651.9

100

653.3

651.9

150

652.32

651.9

considered without constraints and with the aforementioned set of iteration number and population size since the statistical tests reveal there is no significant difference in their performance at these values (as is discussed in Sect. 150.3).

Fig. 3. Graph of best cost recorded v/s the iteration number for GA

Fig. 4. Graph of best cost recorded v/s the iteration number for DE

The graphs above demonstrate that both GA and DE converge to their respective optimal solutions after around 50 algorithm runs. Table 2 indicates the average time required by each algorithm to reach its optimal solution when utilizing the cost function without constraints. We note that, at (population size/maximum number of iterations) = (150/150), DE takes on average less time to reach its optimal value (0.5103 s) than GA (1.0513 s). Despite comparing the time needed by DE to achieve its optimal solution at (150/150) with the population size and maximum number of iterations at which

Optimizing Headways Using Evolutionary Algorithms

73

GA produces its lowest recorded value, specifically at (100/100), we can still conclude that DE required less time than GA. Hence, when considering the cost function without constraint, we can conclude that, while DE may require a larger iteration number and population size for its performance to be deemed statistically equivalent to GA, it produces optimal results within a smaller timeframe than GA. Table 2. Best average costs and computation time for DE and GA (without constraint) Pop. size/max. iteration

DE Best average cost (std. dev)

Average time taken(s)

GA Best average cost (std. dev)

Average time taken(s)

50/100

642.6962 (0.9987)

0.1806

641.4869 (7.2832e−04)

0.4962

100/100

641.6792 (0.3558)

0.3662

641.4868 (4.5936e−13)

0.7874

100/150

641.6283 (0.3731)

0.3500

641.4868 (4.5936e−13)

0.7839

150/150

641.5093 (0.1034)

0.5103

641.4868 (4.5936e−13)

1.0513

3.3 Robustness To determine which algorithm is more robust, we tested both DE and GA with different cost functions. The first cost function has constraints, and the second one does not have any constraints. When tested with the normal cost function (with constraints), both DE and GA algorithms were able to find the minimum value, but with variations in their mean values. The mean values of the cost function for DE were higher than that of GA in all the four tests conducted. The t-tests showed that the means were significantly different. However, when tested with the cost function without constraints (using results obtained from 50 runs of both algorithms), the difference between the mean best values of DE and GA reduced significantly (Table 2). In this case, the mean values for GA and DE were almost the same, with DE having a slightly higher mean value in the first test. We utilized a one-tailed t-test to report the statistical significance of the variance in mean values. The p-values for the initial three tests were below the significance level of 0.05, indicating that there was a notable difference between the average values of both algorithms. In other words, at (50/100), (100/100) and (100/150), GA still generated slightly better solutions than DE. It should be noted that when we increased the maximum iteration number and population size to (150/150), the p-value was calculated to be 0.0647, which is higher than the significance level of 0.05. This outcome implies that the difference between the average values of GA and DE was not statistically significant. Overall, the results suggest that both algorithms are robust and can perform well when tested with different cost functions. When constraints are involved, the DE algorithm

74

O. Gukhool et al.

may have a higher cost value than the GA algorithm, and this difference is statistically significant. However, when constraints are removed, the performance of both algorithms are fairly similar, and the choice between them may depend on other factors such as computation time and ease of implementation.

4 Conclusion In conclusion, this study compared the efficacy of GA and DE algorithms in optimizing headways for LRVs with the goal of minimizing passenger waiting time and operational costs for LRT systems. The results of this study show that both DE and GA algorithms, when applied to the proposed optimization model, are effective in reducing the generalized costs of an LRT network. The optimization algorithms were successful in achieving a decrease of 6.09% and 6.15% in the generalized cost of the LRT network, demonstrating the efficacy of the model and optimization algorithms. The results also showed that GA outperformed DE in finding the optimal solution, as evidenced by the lower mean value of the best solution for GA in all tests. However, DE was found to be faster than GA in completing the optimization process. Both algorithms were found to be robust and able to perform well when tested with different cost functions. Our findings differ from the research described at the start of this paper, since they focused mainly on GA as the optimization algorithm for transit systems other than LRT. Our study compared the performance of GA and DE algorithms in optimizing LRV headways, which is a novel approach. Additionally, our study provides evidence that GA outperforms DE in finding the optimal solution for LRV headways. Our study also provides additional evidence that DE is a viable alternative to GA in this optimization problem and may have advantages in terms of finding the optimal solution. In general, our results indicate that the GA is more efficient than DE for optimizing LRV headways, particularly when considering cost function with constraints. However, the selection of DE or GA may depend on additional factors like computation time and implementation difficulty. Further studies could explore the efficacy of these algorithms in other optimization problems related to LRT systems. Acknowledgement. We are grateful to the Higher Education Commission (Award Number TB2019-08) and all our collaborators, including the University of Mauritius, project investigators, research assistants and training project assistants, Metro Express Ltd, Ministry of Land Transport and Light Rail, Traffic Management and Road Safety Unit, and Road Development Authorities for their invaluable support of this research project.

References 1. Niu, H., Zhou, X., Gao, R.: Train scheduling for minimizing passenger waiting time with time-dependent demand and skip-stop patterns: Nonlinear integer programming models with linear constraints. Trans. Res. Part B: Methodol. 76, 117–135 (2015) 2. Besinovic, N., et al.: Artificial Intelligence in railway transport: taxonomy, regulations, and applications. IEEE Trans. Intell. Transp. Syst. 23(9), 14011–14024 (2022)

Optimizing Headways Using Evolutionary Algorithms

75

3. Zhao, L., Chien, S.I.: Investigating the impact of stochastic vehicle arrivals to optimal stop spacing and headway for a feeder bus route. J. Adv. Transp. 49(3), 341–357 (2014) 4. Thuan, N.Q., Thang, P.N.: A novel model for BRT scheduling problems. Int. J. Mach. Learn. Comput. 9(4), 401–406 (2019) 5. Wang, J.Y.T., Yang, H., Verhoef, E.T.: Strategic interactions of bilateral monopoly on a private highway. Netw. Spat. Econ. 4(2), 203–235 (2004) 6. Xiao, M., Chien, S., Hu, D.: Optimizing coordinated transfer with probabilistic vehicle arrivals and passengers’ walking time. J. Adv. Transp. 50(8), 2306–2322 (2016) 7. Chen, X., et al.: Optimization of headways with stop-skipping control: a case study of bus rapid transit system. J. Adv. Transp. 49(3), 385–401 (2014) 8. Abkowitz, M., Eiger, A., Engelstein, I.: Optimal control of headway variation on transit routes. J. Adv. Transp. 20(1), 73–88 (1986) 9. Liang, S., Ma, M., He, S.: Multiobjective optimal formulations for bus fleet size of public transit under headway-based holding control. J. Adv. Transp. 2019, 1–14 (2019) 10. Schmaranzer, D., Braune, R., Doerner, K.F.: Multi-objective simulation optimization for complex Urban mass rapid transit systems. Ann. Oper. Res. 305(1–2), 449–486 (2019) 11. Mutlu, M.M., Aksoy, ˙I.C., Alver, Y.: Transit frequency optimization in bi-modal networks using differential evolution algorithm. Teknik Dergi (2022) 12. Qian, X., He, J., Liu, Y., Wu, C.: A differential evolution algorithm for multi-objective optimization problems. Soft. Comput. 22(8), 2649–2660 (2018) 13. Das, S., Suganthan, P.N.: Differential evolution: a survey of the state-of-the-art. IEEE Trans. Evol. Comput. 15(1), 4–31 (2011) 14. Bowman, L.A., Turnquist, M.: A: Service frequency, schedule reliability and passenger wait times at transit stops. Trans. Res. Part A: Gen. 15(6), 465–471 (1981) 15. Storn, R., Price, K.: Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces. J. Global Optim. 11(4), 341–359 (1997) 16. Heris, M.K.: Practical genetic algorithms in python and MATLAB–video tutorial. https://yar piz.com/632/ypga191215-practical-genetic-algorithms-in-python-and-matlab. Yarpiz (2020)

Improving the Ergonomics of the Master Data Management System Using Annotated Metagraph D. R. Nikolsky1(B) , A. A. Sukhobokov1,2 , and Goryachkin B.S.1 1 Bauman Moscow State Technical University, Baumanskaya 2-ya 5, 105005 Moscow, Russia

[email protected] 2 SAP CBS, Dietmar-Hopp-Allee 16, 69190 Walldorf, Germany

Abstract. The article highlights the importance of human-oriented design of information systems. A data model is proposed and considered for the asset master data management system in the form of a multilayer graph forest based on the nature of the relationships between enterprise assets. The structure of this data model is described. An equivalent replacement of this data model with an annotated metagraph is proposed. A typical analysis problem solved by the user of such a system is highlighted and ergonomic problems that need to be solved are stated. A solution to these problems is proposed using the entity-resolution algorithm and a decision is made about the ergonomic failure of this solution. An approach is proposed to modernize the user interface by replacing a multilayer graph forest with an annotated metagraph. The ergonomic improvement of the user interface and its positive impact on usability are described. Keywords: Ergonomics · Human-oriented design · Master data management · Multilayer graph forest · Annotated metagraph

1 Introduction With the transition to comprehensive automation of production, the role of humans as the subject of labor and management has increased. Humans are responsible for the efficient operation of the entire technical system, making human-centered design of interactive systems especially important [1]. Ergonomics, specifically microergonomics, deals with the study and design of manmachine systems. It solves the problem of optimizing human functioning in the control circuit of automated information processing and control systems. This includes creating optimal working conditions that increase productivity, provide convenience, and preserve the strength, health, and efficiency of humans.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 76–83, 2024. https://doi.org/10.1007/978-3-031-50158-6_8

Improving the Ergonomics of the Master Data Management

77

2 Information Model of the Proposed System, Description of the Subject Area Master Data Management (MDM) is a set of best practices for managing data that helps key stakeholders, participants, and business customers implement business applications, information management practices, and data management tools. These practices, policies, procedures, services, and infrastructure support the collection, integration, and sharing of accurate, timely, consistent, and complete master data [2, 3]. In these systems, humans directly interact with data to analyze, interpret, and reveal hidden patterns. In this case, the level of organization of the considered ergatic system is the third - the system provides the energy and information functions, and the person is the controlling one. The tasks for which master data management systems are used define the possible mappings of objects with which businesses interact and their relationships, including hierarchical ones [4]. All of an enterprise’s assets are reflected in different areas of the enterprise’s internal operations and accounting. One such area that reflects a holistic view of data from the perspective of one or more interrelated applications is called a perspective. An example of this relationship can be observed in Fig. 1. This means that all information about assets can be represented as a large, complex graph, where the nodes are the enterprise’s assets (i.e., their representations in different perspectives), and the edges are the relationships between their representations [5].

Fig. 1. Example of a lattice of inter-perspective connections

78

D. R. Nikolsky et al.

Each perspective is described by a hierarchical structure, and the structure of the hierarchy itself can vary significantly from one perspective to another. The graph is a set of tree perspectives, which represent the enterprise assets in different ways. An example hierarchy is shown in Fig. 2.

Fig. 2. Example of hierarchical relationships in the perspective

Data hierarchies are very important for MDM-systems, because they provide transparency in the structure of company assets and allow you to clarify the relationship between the master data. Data hierarchies are crucial for MDM systems because they provide transparency in the structure of company assets and clarify the relationship between master data. Connecting the perspectives is necessary to create a unified, holistic view of the master data assets. In this case, an enterprise asset is represented by a grid of nodes in different perspectives. Therefore, the proposed graph is also a set of rings (lattices) between nodes from different perspectives, representing the enterprise assets. An example of such links is shown in the previously discussed Fig. 1. A similar data structure, a multilayer graph forest, can also be represented as an annotated metagraph [6]. Nodes of different perspectives, which are connected by lattices and represent nodes meaning one object, can be enclosed in one meta-vertex to avoid data redundancy in the system. An example of semantically equivalent data structures is shown in Fig. 3.

3 Problem Statement The volume of perception of the human visual analyzer is determined by the number of objects a person can grasp and remember during one visual fixation. When unrelated objects are presented, the volume of visual perception is limited to 4–8 elements. Note

Improving the Ergonomics of the Master Data Management

79

Fig. 3. On the left is a graph structure of data with a lattice of inter-route connections. On the right is an equivalent metagraph.

that the amount of information that can be reproduced is determined not by the amount of perception, but by the amount of memory. A visual image may reflect a larger number of objects, but they cannot be reproduced due to limited memory capacity. Therefore, it is more important to consider memory capacity than perception capacity for practical purposes. To work efficiently, it is necessary that no more than 6 ± 2 elements fall into the central field of vision, limited by an angle of 4–10° for operators [7]. In any system involving an operator, several fundamental principles must be followed. Three of them are of interest in this situation: 1. The principle of maximum understanding. The system must provide full support to the person, with information that does not require interpretation or recoding. 2. The principle of the minimum amount of users’ working memory. A person should remember as little as possible. 3. The principle of optimal loading. It recommends a distribution of functions in which the operator would not experience starvation (loss of activity) or overload (missing signals) at the rate of incoming data. In the proposed MDM-system, when analyzing objects on the screen, the user solves the problem of reduction - identifying the necessary nodes in the graph related to one object in the real world and mentally combining them to highlight the object. Figure 4 shows the layout of the master data management system’s object view screen interface as seen by the user. As the number of inter-view links increases, the lattice of nodes corresponding to one real-world entity creates difficulties for perception.

4 Proposed Solution To ensure high-quality software, the dialog stages must conform to the goal of successfully executing a production task. The dialog should only include necessary stages without which it is impossible to solve the task. Stages that include actions more appropriate for automatic execution by the system must be excluded [8]. To facilitate the user’s task of analyzing objects on the screen, a union of nodes corresponding to a single realworld object must be constructed into a special set that reflects the object. Figure 5 shows the resulting screen layout. Instead of analyzing multiple nodes and graph edges, the

80

D. R. Nikolsky et al.

Fig. 4. The mockup of the data display area of the screen with highlighted inter-perspective connections, reflecting the correspondence of nodes from different perspectives to the same object in the real world.

user only needs to perceive one object displayed by an ellipse, which corresponds to one real-world entity and includes all nodes displaying this object on different perspectives of master data. To search for such objects in the entire data set, the proposed algorithm works as follows: 1. Iterate over every node in all hierarchies. 2. For each node, view and trace its inter-perspective connections and the rings of connections formed by them during traversal. 3. Mark nodes of different perspectives enclosed in one or more rings as one entity. 4. Combine the intersecting and nested sets of nodes found in the previous step into a single entity. The set of such objects forms the user representation shown in Fig. 5. This algorithm allows the user to perform their tasks by creating an aggregated representation of the data. However, due to the nature of the graph structure, the algorithm must be executed every time the user requests the data for display. This is because the graph structure doesn’t allow for storing pre-built indexes. As a result, the principle of optimal loading will be violated, and the user will have to wait for the algorithm to finish execution to get the data presentation. To solve this problem, the data structure itself could be modernized. The introduction of an annotated metagraph as a data structure storing information about assets accomplishes several goals simultaneously:

Improving the Ergonomics of the Master Data Management

81

Fig. 5. Layout of the data display area, modified by bundling nodes.

1. The Real Object Index is built initially, which provides users with an ergonomically correct display of asset data without the need for special processing. 2. The size and volume of the data are reduced to lower the likelihood of errors and incorrectness. This is done by excluding individual inter-perspective links and replacing them with meta-vertices. An annotated metagraph is chosen from various types of metagraphs because of the presence of meta-vertices, which may include other vertices and metavertices as their internal elements. The choice of an annotated metagraph from various types of metagraphs is due to the presence of metavertices, which may include other vertices and metavertices as their internal elements. If asset information is presented using an annotated metagraph, the user will see the screen layout shown in Fig. 6. It can be seen that the inter-perspective connections that take away the user’s focus are abolished and replaced by meta-vertices. In this case, the semantic completeness of the data is not lost because the user can highlight both individual objects of the real world in the system’s information field and specific representations of these objects in different perspectives of the enterprise’s business activities.

82

D. R. Nikolsky et al.

Fig. 6. Layout of the data display area. The system uses a metagraph as a data structure to store asset information.

5 Conclusion Using an annotated metagraph-based data model to represent master data on assets does not affect user performance because the semantic correctness of the stored and displayed data is not changed. At the same time, this model provides the following advantages: 1. The efficiency of user activity is increased by reducing the time and mental resources spent by the user to solve the problem. 2. User satisfaction increases by reducing mental overload and removing the task of entity resolution from the user. A change in these indicators leads to a cumulative improvement in the usability of the system [9].

References 1. ISO 9241-210. https://www.iso.org/standard/77520.html. Last accessed 10 Feb 2023 2. Loshin, D.: Master data management. Morgan Kaufmann (2010) 3. Talburt J., Zhou Y.: Entity information life cycle for big data: Master data management and information integration. Morgan Kaufmann (2015) 4. Ng, S.T., Xu, F.J., Yang, Y., Lu, M.: A master data management solution to unlock the value of big infrastructure data for smart, sustainable and resilient city planning. Proc Eng 196, 939–947 (2017) 5. Sukhobokov, A.A., Strogonova, V.I.: On an approach to construct asset master data management system. Softw Syst 30(1), 51–60 (2017)

Improving the Ergonomics of the Master Data Management

83

6. Gapanyuk, Y.: The development of the metagraph data and knowledge model. In: Selected contributions to the 10th international conference on “integrated models and soft computing in artificial intelligence (IMSC-2021). CEUR WORKSHOP PROCEEDINGS, vol. 2965, pp. 1–7 (2021) 7. Goryachkin, B.S.: Ergonomic passport of an automated system for processing and displaying information and control. Int Res J 9–2(51), 25–29 (2016) 8. Goryachkin, B.S., Umryadev, D.T.: The role of software ergonomics standards in the analysis, design and evaluation of information systems software. Trends Develop Sci Educ 73–3, 153– 161 (2021) 9. ISO 9241–11. https://www.iso.org/standard/63500.html. Last accessed 27 Jan 2023 10. Intelligent Computing & Optimization, Conference proceedings ICO 2018, Springer, Cham, ISBN 978-3-030-00978-6 11. Intelligent Computing and Optimization Proceedings of the 3rd International Conference on Intelligent Computing and Optimization 2020 (ICO 2020) 12. Intelligent Computing & Optimization Proceedings of the 4th International Conference on Intelligent Computing and Optimization 2021 (ICO2021) 13. Intelligent Computing & Optimization Proceedings of the 5th International Conference on Intelligent Computing and Optimization 2022 (ICO2022) 14. Special Issue. https://link.springer.com/journal/40305/volumes-and-issues/10-4

Patent Classification for Business Strategy with BERT Masaki Higashi1(B) , Yoshimasa Utsumi2 , and Kazuhide Nakata1 1 Tokyo Institute of Technology, Tokyo, Japan [email protected], [email protected] 2 Rakuten Group, Inc., Tokyo, Japan [email protected]

Abstract. Recent developments in deep learning have greatly improved the accuracy of various natural language processing tasks. Patent classification using patent documents has also seen a significant improvement in accuracy based on the internationally defined IPC as well as the region-specific CPC prediction methods. Such an automated patent classification capability can reduce the burden on examiners and improve efficiency. It can also be used to organize patents, such as in prior art searches. However, such a classification cannot be used to support the formulation of management strategies, such as capturing the fields in which the predicted patent classification will be used in the management of a company. To this end, this study aims to classify patents based on technological field instead of the existing IPC classification such that the developed automated patent classification is useful for companies. To elaborate, this study investigates the differences between IPC and the proposed classification based on technological field, selects documents to be used, and proposes a classification model using IPC information. The proposed classification model is comprised of two distinct components, namely a model for document input and another for IPC input. BERT is employed for the document input model, while skip-connections are used for the IPC input model. Finally, the improvement in accuracy is examined compared to existing IPC using actual data; moreover, solutions are proposed to the problems identified in this study. As a result, our proposed model significantly improves precision, recall, F1, and AP over existing models. Keywords: Deep learning · BERT · Patent classification · Graph embedding

1 Introduction With the development of deep learning, the accuracy of machine learning using document data has been greatly improved. In particular, the Transformer model [12] has contributed greatly to improving the accuracy of various tasks regarding document data. Furthermore, the emergence of BERT [2], which performs pre-training using the Transformer model, has greatly improved the accuracy of natural language processing tasks, allowing realworld applications. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 84–94, 2024. https://doi.org/10.1007/978-3-031-50158-6_9

Patent Classification for Business Strategy with BERT

85

Currently, the main patent classification methods include the internationally defined IPC and the region-specific CPC. Since these are assigned manually by humans, cost and human error are major issues. Therefore, deep learning with patent documents is used to improve the accuracy of patent classification. For example, Patent BERT, which uses BERT as described earlier, has succeeded in improving the accuracy beyond aforementioned methods using deep learning. However, most of the current patent classification methods focus on IPC that can be used to organize patents but cannot be used for management strategies of companies that own patents or are going to use patents. This is because the IPC is classified into categories in which the uses and characteristics of patents are briefly described, and many specific uses are possible from a single IPC even if they are applied to management strategies or services as they are; for example, “G08G 5/00” mean “traffic control system for aircraft”. It is difficult to consider specific management strategies and services from the IPC alone, because “traffic control system for aircraft” has various possible applications, such as “automatic piloting”, “air route prediction”, and so on. Therefore, we currently use a search formula that combines information on the type of IPC that a patent has, and the type of words that comprise the text of the patent, and the type of word combinations used, to determine whether a patent is useful for a certain business or not. However, since these are also performed manually on a rule-based basis, cost, human error, and omissions are major issues, as in the case of IPC. Therefore, patents must not only be classified by the IPC but also by more detailed categories, such as “electronic money”, and “e-books”. In this paper, a dataset of categories that can be used for corporate management strategies is prepared using a search formula, and the dataset is applied to an existing method, Patent BERT with “Abstract”. Moreover the “Detailed Description of Invention” section of a patent and IPC information are incorporated into the learning model as a more detailed description than “Abstract” and compared to Patent BERT with “Abstract”. However, two problems emerge. First, the detailed description of the invention is long and includes many contents that are not relevant to classification. Second, the IPC is assigned only as a number and has no specific features. Therefore, a rule-based method is proposed for extracting the necessary contents from the detailed description of the invention, and a method is developed for creating IPC features using a graph structure that represents the IPC sharing relationship among patents. Finally, a model is proposed that combines these methods with summary sentences, and the model is then validated through numerical experiments.

2 About Patents In this section, patents are explained and discussed. First, the IPC is explained, followed by patent documents component. 2.1 IPC When a patent is granted, it is assigned an IPC according to its field. IPC is an internationally unified classification, which consists of four elements: “Section”, “Class”, “Subclass”, and “Main Group/Subgroup”. For example, about “G08G5/00”, “G” in

86

M. Higashi et al.

“Section” means “Physics” and “G08” in “Class” means “Signal”. The “subclass” of “G08G” means “Traffic control system”, and “Sub group/Main group” of “G08G5/00” means “Traffic control system for aircraft”. 2.2 Patent Document The patent document is composed of many elements. In particular,“Abstract”, “Claim”, “Detailed Description of Invention” are the texts containing the description of the patent; moreover, they are important parts of the patent classification using documents. Among them, the “Detailed Description of Invention” is a text item that contains a large amount of information about the patent, and is mainly composed of “Technical field”, “Background technology”, “Prior art document”, “Summary of invention”, and “Brief description of figure”. The “Summary of invention" is the section that describes the patent and comprises “Problem to be solved by the invention", “Means to solve the problem", and “Effects of the invention". The “Problem to be solved by the invention" text item describes the motivation and purpose for developing the patent. Moreover, the “Means to solve the problem" text item describes the proposed approach to achieve the purpose. Finally, the “Effects of the invention" text item describes the implications of the patent. As described in this section, the “Detailed Description of Invention” section provides multiple details of the patent.

3 Related Research As described in the previous section, most of the current patent classifications predict the categories of IPC and other patent classification notations. Example models include Support Vector Machines (SVM) [1, 10] and k-Nearest Neighbor (KNN) [1] models. In addition, there exist models that utilize deep learning techniques, including those employing long-short-term memory (LSTM) [7], convolutional neural networks (CNN) [6], and Bidirectional Encoder Representations from Transformers (BERT) [5] models. Therefore, in this section, BERT model is first introduced that predict IPC, etc.; next, a model is introduced that classifies patents into categories other than IPC. Finally, the relevance of these models to this study is described. Lee et al. [5] proposed PatentBERT. PatentBERT uses the USPTO-2M dataset and the USPTO-3M dataset, which contains 3,050,615 patents with 632 and 656 subclasses for IPC and CPC, respectively. Compared to DeepPatent [6], the accuracy was improved in precision@1 and F1@5. A comparison of the predictability of IPC and CPC using claims shows that CPC is more predictable with regards to precision@1, recall@1, and F1@1. The model proposed by Choi et al. [11] is different from the aforementioned study in that it does not predict IPC, but rather predicts whether the technology under development to avoid patent infringement belongs to a category created by the authors. The language model is Transformer, and the summary sentences are inputted by using word2vec [8] for word embedding. Moreover, features are created using the co-occurrence graphs of IPC, CPC, and USPG and diff2vec [9]. Finally, the output of the Transformer and the created features are used to make predictions. The four categories are “Marine Plant Using Augmented Reality Technology (MPUART)”, “Technology for

Patent Classification for Business Strategy with BERT

87

1MW Dual Frequency System (1MWDFS)”, “Technology for Micro Radar Rain Gauge (MRRG)”, and “Technology for Geostationary Orbit Complex Satellite (GOCS)”. The datasets are retrieved from the USPTO by extracting important keywords from each category. Compared to PatentBERT, “MPUART” and “MRRG” show a decrease in precision but an improvement in recall, F1, and AP. However, “1MWDFS” decreases in precision, recall, F1, and AP, and “GOCS” improves in recall but not in precision, F1, and AP. In this study, IPC features are first extracted with reference to Choi et al. [11] and then compared with those obtained from PatentBERT. In addition, while previous studies used the abstract section of a patent, such as summary sentences, this study uses the “Detailed Description of the Invention” section to distinguish patents more clearly.

4 Proposed Method In this section, first, a method is proposed for extracting sentences of a length that can be inputted into BERT on a rule basis from the “Detailed Description of Invention" section of Japanese patent documents. Next, the IPC feature extraction method is described. Lastly, the proposed model based on BERT and the problems of the proposed model are described. 4.1 Extraction of Documents from Detailed Description Items As mentioned in Sect. 3, in the conventional methods, the title, abstract, and claims are used as features of patent documents. However, the categories used in the proposed classification are more detailed than the conventional IPC. As an example, consider the application number of “Patent Application 2020-524926” that has an IPC of “G08G 5/00”, which implies “Traffic control system for aircraft”. The practical applications of the “Traffic control system for aircraft” classification may include various tasks such as “autopilot”, and “air route prediction”. The technical field assigned to this patent in this research is “logistics means (drone/unmanned delivery)”, which is more practical and detailed than the IPC. This indicates that a patent must be distinguished from other patents more clearly using sentences that include more concrete contents, rather than abstract documents such as titles, abstracts, and claims of existing methods. Therefore, the “Detailed Description of Invention” section of a patent is focused upon in this study. However, note that this section does comprise contents unrelated to the present classification at times, and the lengths of the sentences themselves are very long. Step 1 : Use the items “Problem to be solved by the invention” and “Means for solving the problem” in the detailed description of the invention. Step 2 : If none of the above items in 1 are applicable, use those summarized in “Summary of Invention”, “Overview”, and “Invention Disclosure”. Step 3 : If none of the items in 2 are applicable, a document in which the items after "Brief Description of Drawings" are deleted is used. First, the items in the detailed description of the invention that describe the characteristics of the patent in detail are “Problem to be solved by the invention” and “Means for solving the problem”. However, not all patents have these sections, and many patents that

88

M. Higashi et al.

do not have these items are described using the “Summary of Invention”, “Overview”, and “Invention Disclosure” items, which are then used. Finally, for documents without the “Summary of Invention”, “Overview”, and “Invention Disclosure” items, the documents with the items after “Brief Description of Drawings” deleted are used to excludes sentences that do not describe the patent. 4.2 IPC Feature Generation Methods In conventional patent classification, IPC information cannot be used because the task is to predict the IPC. However, in this study, since classification is performed based on categories other than IPC, IPC can be used. Therefore, IPC features are generated using Node2vec [3] with reference to the co-occurrence graph proposed by Choi et al. [11]. A co-occurrence graph is a graph in which nodes that co-occur with each other are connected by edges. For example, if a patent A exists with IPCs G06F3/06 and G06F13/14 as well as a patent B with IPCs G06F3/06, G06F13/14, and G08G 5/00, then an edge of weight 1 exists between nodes G06F3/06 and G08G5/00; nodes G06F13/14 and G08G 5/00; and nodes G06F3/06 and G06F13/14. Moreover, an edge of weight 2 exists between nodes G06F3/06 and G06F13/14. Next, the features of each IPC are obtained from the co-occurrence graph using Node2vec. Finally, the IPC features are averaged and inputted into the model as the IPC features of the patent. 4.3 Model Structure The model proposed in this section describes the overall picture of a language model that inputs the abstract text of a patent and a detailed description of the invention, of an IPC model that inputs IPC features, and of a model that combines them. This task is considered to be a multi-label classification because a patent may belong to more than one category. Therefore, the loss function is Binary Cross Entropy. The language model uses BERT. The input document is “sent”, and the number of categories is C. Moreover, the linear transformation parameters are W ∈ R768×C , and the label attached to one-hot for each category is y ∈ {0, 1}C . When yi , yˆ i is the i-th element of y, yˆ , the model structure is represented by Eqs. (1)–(3):

Loss = −

x = BERT(sent)

(1)

yˆ = sigmoid(Wx)

(2)



  yi log yˆ i + (1 − yi ) log(1 − yˆ i )

(3)

i∈Z,i∈[1,C]

First, documents are inputted to BERT using Eq. (1), and the number of dimensions of the features is adjusted to the number of categories using a linear transformation in Eq. (2). Subsequently, the documents are passed through a sigmoid function to set the output to (0, 1). Finally, the loss value is outputted using Binary Cross Entropy, as represented in Eq. (3).

Patent Classification for Business Strategy with BERT

89

The model with both the abstract and the detailed description of the invention is expressed in Eqs. (4)–(6): x = BERTabst (abst) + BERTdetail (detail)

(4)

yˆ = sigmoid(Wx)

(5)



Loss = −

  yi log yˆ i + (1 − yi ) log(1 − yˆ i )

(6)

i∈Z,i∈[1,C]

The summary text is inputted as “abst”. Additionally, the detailed descriptions are inputted as “detail”, and the other variables are used in the same way as in Eq. (1). First, the summary and detailed descriptions are entered into different BERTs and then added together using Eq. (4). Next, as in Eq. (5), linear transformations are used to match the number of dimensions of the features to the number of categories. The output is then passed through the sigmoid function to obtain (0, 1). Finally, the loss value is outputted using Binary Cross Entropy. The IPC model is expressed in Eqs. (7)–(9): j

j

j j−1

j−1

xipc = W2 (ReLu(W 1 xipc )) + xipc

(7)

yˆ = sigmoid(W ipc xnipc )

(8)

Loss = −



  yi log yˆ i + (1 − yi ) log(1 − yˆ i )

(9)

i∈Z,i∈[1,C]

For x0ipc using the IPC features generated in Sect. 152.2, the linear transformation parameters are W 1 ∈ Rd ×d , W 2 ∈ Rd ×d , and W ipc ∈ Rd ×C . Moreover, n is the number of layers in Eq. (7). The other variables are used in the same way as in Eqs. (1)–(3). Note that d is the number of dimensions of the IPC features generated in Sect. 152.2, and j is the number of times that Eq. (7) is repeated. Furthermore, the relationship 1 ≤ j ≤ n holds true. First, the IPC features generated in Sect. 152.2 are linearly transformed with W 01 using Eq. (7) when j = 1. Next, they are passed through the activation function ReLu function and then linearly transformed with W 02 . Subsequently, since no information must be lost as the layers get deeper, skip-connection [4] is used to add the before and after transformations using Eq. (7) when j = 1. After this step is repeated n times, a linear transformation is used, as in Eq. (8), to match the dimensionality of the features to the number of categories. The output is then passed through the sigmoid function to set the output to (0, 1). Finally, the loss value is outputted using Binary Cross Entropy. Finally, a model that combines IPC and documents is described. First, on the document model side, when either the abstract or detailed description is used, the document feature obtained by Eq. (1) is used; moreover, when both the abstract and detailed description are used, the document feature obtained by Eqs. (4)–(6) is used. The IPC features obtained by n iterations of Eq. (7) are used as the input for the IPC model. Next, a linear transformation is applied to the concatenated documents and IPC features j

j

90

M. Higashi et al.

to match the dimension of the features to the number of categories, and the output is passed through a sigmoid function to set the output to (0, 1). Finally, the loss values are outputted using Binary Cross Entropy. Figure 1 shows the combined document and IPC model when using both the abstract and detailed descriptions. The model is used to provide a detailed description and abstract; a comparison of the two combined; and a comparison of the two with and without IPC.

Fig. 1. Model with IPC and patent documents

4.4 Problems in the Model and Developed Solutions Choi et al. [11] reported that accuracy is better when IPC is combined with abstract sentences than when abstract sentences are used alone. However, when the accuracy of the model with IPC and abstract sentences as input is compared to the model with the number of layers of the IPC model n = 1 as input, in Sect. 5, the former model showcases no improvement in accuracy compared to the latter model, even though IPC is used as an additional input. First, the major difference between the existing studies [11] and the proposed model are considered. The existing studies used a single-layer Transformer Encoder as the language model, while our model uses a BERT-based model with 12 layers of Transformer Encoders. Therefore, the combination of the IPC and BERT models is considered to be problematic, and the learning speed of each model is investigated. Figure 2 shows the results of this study; note that a large discrepancy exists in the speed of convergence of the learning by the BERT and IPC models. In this situation, BERT model converges first and overlearning occurs, resulting in a convergence of the entire model without a convergence of the IPC model. This diminishes the benefit of including IPC information. However, since BERT is pre-trained on large data sets, the model itself changes if only one layer of the transformer is used, as in existing studies, and the pre-training becomes meaningless. Therefore, the learning speed of the IPC model is increased to make it closer to that of BERT by deepening the layers of the IPC model.By increasing the number of layers n in the IPC model described in Sect. 152.3, the learning speed increases; moreover, a comparison is shown in Fig. 3. The learning speed of the IPC model approaches that of the BERT model when the number of layers is n = 60, thus benefitting from the input of IPC information. A quantitative evaluation of accuracy is described in Sect. 5.

Patent Classification for Business Strategy with BERT

91

Fig. 2. Learning curve of BERT and IPC model

Fig. 3. Learning curve of IPC models

5 Experimental Results This section compares the accuracy with and without the detailed description and abstract sections of a patent as well as discusses the results. Next, the changes in the accuracy with and without IPC are discussed. Finally, the changes in accuracy with and without the solution described in Sect. 152.4 are discussed. 5.1 Experimental Setup A total number of labels is 95 and 8160 patents are assigned multiple labels. The experiments are conducted using training, validation, and test data divided in the ratio of 6:2:2, respectively. The BERT model was fine-tuned, and the parameters of the IPC model were updated using the training data. Subsequently, the validation data were used for inference, and the training process was terminated if the loss value did not show any improvement for five consecutive epochs. Finally, the evaluation value was calculated by performing inference on the test data. For the hyperparameters of the model, Node2vec is fixed as dimensions = 128, walk length = 10, num walks = 10, p = 0.5, q = 2, and window = 3. For the number of layers n of the IPC model, n = 60 is used for both IPC and documents, as described in Sect. 152.4. Furthermore, a pre-trained BERT base model (number of Transformer blocks = 12, hidden layer size = 768, and number of Multi-Head-Attention heads = 12) provided by Tohoku University is used. For the other hyperparameters, a batch size of 32 and a learning rate of 2e-5 are used. We use precision, recall, F1, and AP as the evaluation indices.

92

M. Higashi et al.

5.2 Comparison of the Abstract and Detailed Description Table 1 lists the accuracy of the abstract, the detailed description, and both documents together. Note that the accuracy of the summary is about 0.02 higher than that of the detailed description in all evaluation indices. Moreover, the use of both the detailed description and abstract improves the accuracy by 0.03∼0.04 compared to classification using either the detailed description or abstract. In other words, the abstract contains more information than the detailed description, but adding the information from the detailed description to the case of using only the abstract makes the distinction between patents more accurate. Table 1. Accuracy ratings Precision

Recall

F1

AP

Abstract

0.620

0.612

0.616

0.630

Detailed description

0.604

0.593

0.598

0.613

Abstract & Detailed description

0.657

0.642

0.649

0.676

5.3 Comparison Based on the IPC and Number of Layers Table 2 lists the experimental results for the IPC model with n = 1 and n = 60 layers. Tables 1 and 2 summarize that the accuracy appears to improve with IPC (n = 1) when the detailed description is used, but not for the other models. Tables 1 and 2 summarizes that for abstract, the inclusion of IPC (n = 60) improves precision, recall, and F1 by 0.03 as well as AP by about 0.06. For the detailed description, IPC (n = 60) improves the four evaluation indices by 0.05∼0.06; moreover, for the combination of abstract and detailed description, IPC (n = 60) improves precision, recall, and F1 by about 0.01 as well as AP by about 0.03. Thus, when the number of layers n = 1, the accuracy does not change significantly even when IPC is incorporated. However, by bringing the learning speeds of BERT and IPC models closer together, the accuracy significantly increases with regards to precision, recall, F1, and AP whether the summary and detailed description are by themselves or combined. In other words, when two models with different learning speeds are combined, the models will not be able to learn well if the learning speeds are not aligned.

6 Conclusion and Future Work To classify patent documents based on technology for the purpose of utilizing them in management strategies, this study improves the existing Patent BERT method from the three viewof documents used, IPC feature generation, and model combination. First, the accuracy of classification is successfully improved by incorporating the detailed description section of a patent compared to the existing method of using only

Patent Classification for Business Strategy with BERT

93

Table 2. Accuracy evaluation of IPC model with n = 1 and n = 60 layers n=1

n = 60

Precision

Recall

F1

AP

IPC & abstract

0.625

IPC & detailed description

0.639

IPC, abstract & detailed description

0.654

Precision

Recall

0.619

0.622

0.629

0.634

0.645

0.649

F1

AP

0.633

0.655

0.622

0.662

0.644

0.649

0.690

0.647

0.655

0.677

0.682

0.670

0.654

0.661

0.705

the abstract. Furthermore, IPC information is demonstrably extracted by creating an IPC co-occurrence graph and generating features using Node2vec. In addition, this study demonstrates that combining a model with a large learning convergence speed, such as BERT, with a model having a small convergence speed does not improve accuracy significantly; however, the accuracy improves significantly by bringing the convergence speeds closer together. Three issues need to be addressed in the future. The first is the proportion of labels. The data acquired are generally unbalanced, with a large number of certain labels assigned to a large number of datasets, and a small number of labels assigned to a small number of datasets. This may result in a good prediction for the former label, but not for the latter. However, since labeling is done manually, expanding the dataset is difficult. Therefore, a learning method such as semi-supervised learning must be considered using this model. The second is the use of IPCs. While the IPCs used in this experiment are narrowed down in advance by the search formula, the feature values of a large number of IPCs must be obtained if these IPCs are to be applied to a wide range of fields. In addition, the use of a large number of IPCs may cause noise in the information, making it impossible to extract the information properly. Therefore, experiments must be conducted with more data to see if the proposed model can be adapted to a large number of IPCs, and the approach for extracting features in such scenarios must be investigated. The third is to make better use of the graph structure of patents. Currently, research is being conducted to extract features from the graph structure of papers using Graph Convolution Neural Networks. By applying this method to patents as well, a more accurate classification can be achieved.

References 1. Benzineb, K., Guyot, J.: Automated patent classification. Curr. Challenges Pat. Inf. Retr. 29, 239–261 (2011) 2. Devlin, J., Chang, M.-W., Lee, K., Toutanova, K.: BERT: Pre-training of deep bidirectional transformers for language understanding. In: TProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, vol. 1, pp. 4171–4186 (2019)

94

M. Higashi et al.

3. Grover, A., Leskovec, J.: node2vec: scalable feature learning for networks. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 855–864 (2016) 4. Kaiming, H., Xiangyu, Z., Shaoqing, R., Jian S.: deep residual learning for image recognition. Comput. Vis. Patt. Recog. 770–778 (2016) 5. Lee, J.-S., Hsiang, J.: PatentBERT: Patent classification with fine-tuning a pre-trained BERT model. World Pat. Inf. 61, 101965 (2020) 6. Li, S., Hu, J., Cui, Y., Hu, J.: DeepPatent: patent classification with convolutional neural networks and word embedding. Scientometrics 117(2), 721–744 (2018) 7. Marawan, S., Jan S., Matthias, S., Stephan, G.: An LSTM approach to patent classification based on fixed hierarchy vectors. In: SIAM International Conference on Data Mining 495–503 (2018) 8. Mikolov, T., Sutskever, I., Chen, K., Corrado, G., Dean, J.: Distributed representations of words and phrases and their compositionality. Neural Inf. Process. Syst. 26, 3111–3119 (2013) 9. Rozemberczki, B., Sarkar, R.: Fast sequence-based embedding with diffusion graphs. Int. Workshop Complex Netw. 99–107 (2018) 10. Chih-Hung, W., Yun, K., Tao, H.: Patent classification system using a new hybrid genetic algorithm support vector machine. Appl. Soft Comput. 10, 1164–1177 (2010) 11. Choi, S., Lee, H., Park, E., Choi, S.: Deep learning for patent landscaping using transformer and graph embedding. Technol. Forecast. Soc. Change 175, 121413 (2022) 12. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, Ł., Polosukhin, I.: Attention is all you need. Adv. Neural Inf. Process. Syst. 30 (2017)

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria Decision Making Approach in Rapti River Watershed, India Raashid Khan1 , Jawed Anwar1 , Saif said1 , Sarfarazali Ansari1 , Azazkhan Ibrahimkhan Pathan2(B) , and Lariyah Mohd Sidek3 1 Department of Civil Engineering, Zakir Husain College of Engineering and Technology,

Aligarh Muslim University, Aligarh, India [email protected] 2 Indian Institute of Technology Gandhinagar, Gandhinagar, Gujarat, India [email protected] 3 Institute of Energy Infrastructure, Civil Engineering Department, College of Engineering, Universiti Tenaga Nasional, (The Energy University) UNITEN, Kajang, Malaysia

Abstract. The flood is a catastrophic event that causes losses in life and property. The magnitude of food-related losses has prompted researchers to place a greater emphasis on robust and comprehensive modelling techniques for mitigating food damage. Recently developed multi-criteria decision making (MCDM) techniques are extensively used to make decision-making processes more collaborative, logical, and efficient. The Rapti River is among the most flood-prone rivers in North-Eastern Uttar Pradesh, where numerous villages and towns are annually inundated by monsoon flooding. The present study is undertaken with an aim to identify and map areas of flood hazard and risk in the Rapti River watershed by employing MCDM based analytical hierarchy process (AHP) approach within GIS interface. Flood hazard analysis was carried out by considering eight hazard indicators namely Height above nearest drainage, distance from river, elevation, land use land cover, slope, soil type, drainage density and rainfall. The analysis of flood risk was performed by employing flood hazard layer, population density, and land use land cover as risk indicators. A weighed overlay approach was implemented to prepare flood hazard and risk maps of the study area. The results of the analysis showed that around 57.61% and 62.67% of the study region fall within the moderate category of flood hazard and risk intensity respectively. Furthermore, significantly high i.e., 27.57% and 12.75% of the study region fall into the category of high hazard and risk intensity respectively. The findings of this study suggest that the integration of AHP and GIS techniques provides an effective tool for decision making procedures in flood hazard and risk mapping, as it enables a coherent and effective use of spatial data. Keywords: GIS · AHP · Flood hazard · Flood risk · Rapti River

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 95–103, 2024. https://doi.org/10.1007/978-3-031-50158-6_10

96

R. Khan et al.

1 Introduction Flooding is a recurring hazardous event that results in increasingly global losses of life and property [1]. The Indo-Gangetic and Brahmaputra plains are among the most floodprone areas in India and are considered the worst flood-exaggerated regions in the entire world. Every year, states like Uttar Pradesh, Bihar, and West Bengal situated in the IndoGangetic basin experience severe floods due to the high magnitude of discharge during the monsoon season and the large volume of sediment brought down by the Himalayan Rivers and their tributaries [2]. These floods cause great loss of life and property, and damage to existing infrastructure, such as roads, railways, bridges, and agricultural land. The Rapti River is one of the most flood-prone rivers in North-East Uttar Pradesh. Several villages in the districts of Gorakhpur, Deoria, Maharajganj, Balrampur, and Siddharthnagar get inundated every year due to floods during the monsoon season. The region has experienced four major floods within two decades viz. 1998, 2001, 2007, and 2017. The severity and recurrence of these floods have been predicted to rise over time as a result of climate change [3, 4]. The recently developed and widely accepted MCDM-based AHP method has been employed in several flood hazard and risk assessment studies, revealing fast and precise prediction and is capable of handling inherent complexities resulting from involvement of multiple components of flood causative elements. Pathan et al. [5] employed two decision making techniques i.e., a strategy for order of preference by similarity to ideal solution (TOPSIS) and analytical hierarchy process (AHP) to prioritize sub-watersheds of the Ami River Basin, Uttar Pradesh, to assess the flood risk in the basin. TOPSIS technique revealed better results as compared to AHP. The objective of present study is to create flood hazard and risk maps for the Rapti River watershed by considering eight flood hazard indicators (i.e., height from nearest drainage, distance from river, elevation, land use and land cover, slope, soil type, drainage density and rainfall) in order to identify areas that requires the greatest involvement in the development of risk reduction and mitigation strategies. Population density along with flood hazard spatial layer and rainfall were considered as flood risk indicators for the analysis of risk intensity of floods in the study area. The findings of this study could be used as a decision making tool for effective improvisation of flood control plans and initiatives.

2 Study Area The part of Rapti River watershed lying within the Indian territory is selected as the study area. The study area is located in the Northeastern part of Uttar Pradesh, India and covers an area of about 15,000 km2 . The major land-use covers include buildup area, crops and forests. Geographically, the study area is located between 27˚ 59 25 N to 26˚ 15 57 N latitude and 81˚ 40 50 E to 83˚ 51 11 E longitude. The river emerges from the Himalayan range in the Nepalese territory, then enters the plainlands and finally joins Ghagra River which is one of the major tributaries of the Ganga River. The total main stream length of the river within India is about 640 km (Fig. 1).

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria

97

Fig. 1. Location of the study area

3 Methodology The flood hazard and risk analysis was carried out using Geographic Information Systems (GIS) based MCDM approach. The flood hazard and risk indicators were selected and assigned ranks based on an extensive review of the published literature and by collecting the expert’s opinion working in the field of flood modelling and analysis through questionnaires. Height above nearest drainage (HAND), elevation, distance from river, land use (LULC), slope, drainage density, soil, and rainfall were considered as eight important flooding hazard indicators. Analysis of flooding risk was based on flood hazard spatial layer, population density and LULC. Weighted linear combination (WLC) method was used for overlaying different layers to obtain flood hazard and flood risk maps of the study area. SRTM digital elevation model (DEM) of 30 m spatial resolution was download from United States Geological Survey (USGS) web portal and utilized for creating elevation, slope, drainage density and HAND layers within GIS framework. Landsat 8 OLI/TIRS image of October 2020 with 30 m resolution was downloaded from USGS web portal utilized for generating LULC map of the study area. The soil data was downloaded from the National Bureau of Soil Survey and Land Use Planning (NBSS & LUP) and transformed into soil map of the study area. Rainfall data of ten years duration i.e., from 2010 to 2020 was procured from the World Bank Group (climate change portal) and the same were utilized for creating rainfall map of the study area. The comprehensive database pertaining to demographic statistics i.e., population density data was collected from the District Census Handbook (DCH) of the State, and transformed into spatial map layers utilized for flood risk analysis. 3.1 Flood Hazard Indicators Height Above Nearest Drainage (HAND) The relative height of a location in relation to the nearest river tributary is important in determining its flood susceptibility since lowlying lands adjacent to streams are more vulnerable to flooding than higher elevation land.

98

R. Khan et al.

For creating the HAND map, SRTM DEM and the Stream ordering layer are required as inputs. Distance from River (DR) is another important parameter in determining an area’s flood susceptibility. Floods have a greater impact on areas near rivers than on areas farther away from them. The reclassification of the layer was based on assigning a value of 5 to areas farthest from the river, and a value of 1 to areas near to the river. Elevation (E) Elevation of an area has an important role in controlling the movement of the runoff water and the depth of the water level. The Elevation raster map was created using the SRTM DEM and slope generation tools in ArcGIS software. The DEM reveals the southeast region to be more susceptible to flooding owing to relatively low elevation values. Land Use Land Cover (LULC) Urban expansion is considered one of the major contributors to the increased frequency of floods observed in recent times. As impervious cover increases and forest cover shrinks in urban areas, run-off increases. The change in land use directly effects the flooding and susceptibility to damages. Accordingly, water bodies were ranked with the value of 5 as this LULC classification is highly susceptible to floods. Grassland/bare ground were assigned the value of 2, since both land cover features accounts for low susceptibility to floods. Likewise, trees were assigned the lowest rank as 1, due to low susceptibility to flooding. Slope (S) The slope of a feature is its angle of inclination with respect to the horizontal plane. Slope is a key indicator of regions that are highly prone to flooding. The rate and duration of runoff are highly influenced by slope. The runoff accumulates at a much faster rate in flat areas thereby, making these areas more prone to flooding than steeper surfaces. Soil Type (ST ) Soil properties in a watershed, such as permeability, soil layer thickness, rate of infiltration and antecedent soil moisture may have a positive impact on rainfall-runoff process and may eventually lead to flooding of an area under extreme circumstances. Flooding becomes more likely as soil infiltration capacity decreases, resulting in an increased surface runoff. When surface runoff is generated in a quantity that exceeds the soil’s infiltration capacity, it travels down the slope or accumulates on a flat terrain and may lead to flooding. Drainage Density (DD) is the measurement of sum of the drainage lengths per unit basin area and is considered as one of the key factors that influences peak flows during rainfall. Drainage density of an area indicates the nature of soil and its geotechnical properties and therefore, drainage density depends on soil permeability and erodibility, LULC and slope. Drainage density is an inverse function of infiltration. Areas with high drainage density indicates higher runoff, and low flood risk. Thus, the ranking for drainage density decreases with increasing drainage density. Rainfall (RF) The worldwide frequency of severe rainfall events has enhanced and the intensity of flood is directly proportional to the intensity of rainfall. The increase in both severity and quantum of rainfall owing to climate change has significantly brought a rise in severe flooding events. Spatially distributed maps of the eight selected flood hazard indicators and a population density map as flood risk indicator are shown in Fig. 2. For the purpose of generating flood indices using the spatial analyst tool in ArcGIS, the MCDM-based AHP technique determined the relative weights of each raster layer and the spatially distributed flood hazard index (FHI) and flood risk index (FRI) maps were generated using standard expression implying sum product of relative weight and the corresponding raster layer.

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria

(a)

99

(b)

(c)

(d)

(e)

(f)

Fig. 2. Flood hazard Indicators (a) height above the nearest drainage; (b) distance from the tributaries; (c) elevation; (d) LULC; (e) slope; (f) soil classification; (g) drainage density; (h) rainfall; (i) population density

3.2 Analytical Hierarchy Process Saaty [6] developed the Analytic Hierarchy Process (AHP) to handle complicated issues with multiple criteria. This method analyses judgements with the use of mathematical procedures that take into account the preferences of decision-makers or groups of people in a specific field based on selected factors. Nearest neighbor comparisons are performed in this method to establish relative priorities among the multiple factors or criteria involved, and these pairwise comparisons are done on a nine-level standardized

100

(g)

R. Khan et al.

(h)

(i) Fig. 2. (continued)

scale [7, 8]. The relative importance of parameters is measured on a scale from 1 to 9, with 1 indicating that the two parameters are of equal importance and 9 indicating that one parameter is significantly more significant than the other. The reciprocal of Saaty’s rating scale (i.e., 1/1 to 1/9) indicates that one parameter is less important than the other (Table 1). The parameters under consideration are ranked based on their relative importance, which is determined by expert judgement via surveys and Pearson’s inter-correlation between the parameters. The comparison matrix is generated by comparing each parameter one to one, yielding a total of nC2 comparisons. If the judgement criteria fills the upper portion of the diagonal in the comparison matrix, the reciprocals fill the lower portion of the matrix, and the comparison matrix is constructed. The pair wise comparison matrix is then linearly normalized, where each element is split by the sum of the elements in its corresponding column after the sum of all the values from each column has been obtained. The average of all the elements in each row of the normalized matrix is then calculated to determine the relative weights of each factor as seen in Table 2. To check the preciseness of the relative importance of each criteria obtained through the process, the consistency ratio (CR) is applied which is defined as the ratio of consistency index (CI) to the random index (RI). The CR value less than 10% suggest that the judgement on relative importance among criterion/parameters is consistent.

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria

101

Table 1. Saaty’s scale Intensity

Preference degree

Explanation

1

Equally

Both parameters have equal importance

3

Moderately

One parameter is favored slightly more than another

5

Strongly

One parameter is favored strongly than another

7

Very strongly

One parameter is favored very strongly and is considered superior to another

9

Extremely

One parameter is favored as superior to other parameter in highest possible order of affirmation

2,4,6,8 can be used for intermediate of the above values

Table 2. Normalization of the comparison matrix of the flood hazard indicators Flood hazard indicators

HAND

DR

LULC

S

ST

RF

DD

E

Weight

HAND

0.32

0.41

0.33

0.30

0.27

0.23

0.17

0.17

0.27

DR

0.16

0.20

0.33

0.20

0.20

0.19

0.17

0.17

0.20

LULC

0.16

0.10

0.16

0.30

0.20

0.23

0.24

0.19

0.20

S

0.11

0.10

0.05

0.10

0.20

0.14

0.17

0.14

0.13

ST

0.08

0.07

0.05

0.03

0.07

0.14

0.10

0.12

0.08

RF

0.06

0.05

0.03

0.03

0.02

0.05

0.10

0.12

0.06

DD

0.06

0.04

0.02

0.02

0.02

0.02

0.03

0.07

0.04

E

0.04

0.03

0.02

0.02

0.01

0.01

0.01

0.02

0.02

Total

1.00

1.00

1.00

1.00

1.00

1.00

1.00

1.00

1.00

4 Results The consistency ratio (CR) for the flood hazard was evaluated as 8.29% and the principal eigenvalue as 7.67 on the basis of the weight scale derived from the standardized pairwise comparison matrix. For the flood risk evaluation, the CR was computed as 9.82% and the principal eigenvalue 7.78 which shows consistency in the judgement process. The MCDM based AHP approach calculated the weights of various contributing hazard and risk indicators that were multiplied with the spatial layer of respective indicator and overlaid using spatial analyst’s weighted linear combination (WLC) method within the GIS interface. The final formulated expression for FHI is written below and the expression was used for generating the flood hazard map (Fig. 3a). The flood hazard and risk maps were further categorized into 5 classes: i.e., very high, high, moderate, low, and very low.

102

R. Khan et al.

FHI = 0.27 × HAND + 0.20 × DR + 0.20 × LULC + 0.13 × S + 0.08 × ST + 0.06 × RF + 0.04 × DD + 0.02 × E

Fig. 3. (a) Flood hazard and (b) flood risk map of the Rapti River watershed

The population density layer, LULC layer, and the flood hazard layer i.e., FHI, multiplied by considering weights in equal proportion (i.e., 0.33) were utilized to generate the flood risk map of the study area (Fig. 3b). The final formulated FRI expression is written below. FRI = 0.33 × FHI + 0.33 × LULC + 0.33 × Population density The flood hazard map of Rapti River watershed shown in Fig. 3a, reveals that 0.13% (19.76 km2 ), 14.34% (2149.5 km2 ), 57.61% (8634.8 km2 ), 27.57% (4131.6 km2 ), and 0.35% (52.5 km2 ) of the watershed area falls within very low, low, moderate, high, and very high category of flood hazard respectively. Results further reveal that a substantial proportion of the study area lies within moderate to high risk of flooding. The flood risk map for the watershed (Fig. 3b) shows that 5.67% (844.73 km2 ), 18.65% (2780.55 km2 ), 62.67% (9342.19 km2 ), 12.75% (1900.47 km2 ), and 0.26% (38.29 km2 ) of the watershed’s area is at the risk of very low, low, moderate, high and very high flooding respectively. Results are indicative of the fact that a significant proportion of the study area i.e., 85.18% and 75.42% is located in regions with moderate to high flood hazard and risk intensities respectively.

5 Conclusion The most frequent natural disaster that affects people and results in significant loss of life and property worldwide is flooding which occur for numerous reasons and in various situations. Recent reports indicate that the frequency of flood have increased

GIS Based Flood Hazard and Risk Assessment Using Multi Criteria

103

many fold in recent times necessitating precise identification of flood prone regions to ensure the long-term sustainable solutions of flood risk reduction by recommending effective flood control measures on priority basis. This study used MCDM based AHP approach to create flood hazard and risk maps of the Rapti River watershed within GIS interface. A total of eight flood hazard indicators including height from nearest drainage, distance from river, elevation, LULC, slope, soil type, drainage density and rainfall were considered for evaluating the flood hazard. Population density along with FHI spatial layer and cumulative rainfall were considered for evaluating the flood risk in the Rapti River watershed. The degree of food hazard and risk intensity in terms of areal coverage were categorised into five classes: very low, low, moderate, high, and very high. The results of the analysis reveal that almost 57.61% and 62.67% of the study area falls under the moderate category of flood hazard and risk intensity respectively. Moreover, around 27.57% and 12.75% of the study area falls under the high category of hazard and risk intensity respectively. The areas that fall into the moderate and high hazard and risk intensity categories need to have effective flood control measures, such as levees or flood walls along key river sections. It is also important to identify feasible locations to construct flood control reservoirs and rehabilitate storm water drainage systems. The results can be used by disaster management authorities to implement effective and sustainable flood management strategies in flood-prone areas.

References 1. Ogato, G.S., Bantider, A., Abebe, K., Geneletti, D.: Geographic information system (GIS)Based multicriteria analysis of flooding hazard and risk in Ambo Town and its watershed, West shoa zone, oromia regional State, Ethiopia. J. Hydrol. Reg. Studies 27 (2020). https:// doi.org/10.1016/j.ejrh.2019.100659 2. Pathan, A.I., Agnihotri, P.G., Patel, D.: Integrated approach of AHP and TOPSIS (MCDM) techniques with GIS for dam site suitability mapping: a case study of Navsari City, Gujarat. India. Environ. Earth Sci. 81(18), 443 (2022) 3. Dewan, A., Islam, M.M., Kumamoto, T., Nishigaki, M.: Evaluating flood hazard for land-use planning in greater Dhaka of Bangladesh using remote sensing and GIS techniques. Water Resour. Manage. 21, 1601–1612 (2006). https://doi.org/10.1007/s11269-006-9116-1 4. Abah, R.C.: An application of geographic information system in mapping flood risk zones in a north central city in Nigeria. African J. Environ. Sci. Technol. 7, 365–371. https://doi.org/10. 5897/AJEST12.182 5. Pathan, A., Kantamaneni, K., Agnihotri, P., Patel, D., Said, S., Singh, S.K.: Integrated flood risk management approach using mesh grid stability and hydrodynamic model. Sustainability 14(24), 16401 (2022) 6. Saaty, T.L.: A scaling method for priorities in hierarchical structures. J. Mathem. 15(3), 234–81 (1977) 7. Uddin, K., Gurung, D.R., Giriraj, A., Shrestha, B.: Application of remote sensing and gis for flood hazard management: a case study from sindh province, Pakistan. Am. J. Geogr. Inf. Syst. 2013, 1–5 (2013). https://doi.org/10.5923/j.ajgis.20130201.01 8. Pathan, A.I., Girish Agnihotri, P., Said, S., Patel, D.: AHP and TOPSIS based flood risk assessment-a case study of the Navsari City, Gujarat. India. Environ. Monit. Assess. 194(7), 509 (2022)

Optimizing Laser Drilling of Kenaf/HDPE Composites: A Novel CRITIC-MABAC Methodology Sellamuthu Prabhukumar1 , Jasgurpeet Singh Chohan2 , and Kanak Kalita3(B) 1 Department of Mechanical Engineering, Presidency University, Bangalore, India

[email protected]

2 Department of Mechanical Engineering and University Centre for Research & Development,

Chandigarh University, Mohali, India 3 Department of Mechanical Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of

Science and Technology, Avadi, India [email protected]

Abstract. In this paper, a novel hybrid multi-criteria decision-making (MCDM) methodology called the CRITIC-MABAC approach is introduced. It is applied in optimization of laser drilling of kenaf/high-density polyethylene (HDPE) composites. The objective is to simultaneously minimize the kerf taper angle (ϕ) and surface roughness (Ra ) by optimizing the two process parameters (namely laser power and cutting speed). To establish the efficacy of the CRITIC-MABAC, its results are compared with Entropy-MABAC approach. The optimal parametric combination is found to be at a laser power of 120W and cutting speed of 4 mm/s. Thus, the proposed CRITIC-MABAC method demonstrates its effectiveness and can be used as a means to achieve superior hole quality. Keywords: Composites · Laser drilling · CRITIC · MABAC · Multi-criteria decision-making · Optimization

1 Introduction Composites have become widely popular in the recent years due to their eco-friendliness, low cost and superior mechanical properties. They are widely used in numerous industries like automotive, construction, and packaging. Achieving high-quality hole characteristics is crucial for the assembly and functionality of composites, making the drilling process an essential operation in their manufacturing. Laser drilling is a non-contact, highly precise, and efficient technique for creating holes in a variety of materials. A focused laser beam is used to remove material by vaporization and melting, resulting in the formation of a hole. However, achieving the desired hole quality characteristics in kenaf/high-density polyethylene (HDPE) composites can be challenging due to the heterogeneous nature of the composite and the sensitivity of the kenaf fibers to thermal degradation. Therefore, it is vital to optimize the laser drilling © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 104–111, 2024. https://doi.org/10.1007/978-3-031-50158-6_11

Optimizing Laser Drilling of Kenaf/HDPE Composites

105

process variables to realize least kerf taper angle (ϕ) and surface roughness (Ra ) while maintaining the structural integrity of the composite. In the literature, various studies have conducted the optimization of laser drilling processes for different materials. Singh et al. [1] conducted a comprehensive study on the laser drilling of HDPE. They employed ANOVA and response surface methodology for optimization. The cutting speed and laser power were considered as the process variables for optimizing Ra , ϕ, and heat-affected zones (HAZ). Researchers have also investigated the optimization of mechanical properties and surface quality in laser-drilled glass fiber-reinforced plastic (GFRP) laminates [2]. Several studies have explored conventional drilling operations, such as the work by Palanikumar et al. [3], who investigated the drilling of GFRP using an L9 orthogonal array. They employed Grey Relational Analysis (GRA), an MCDM technique, to optimize the drilling process. Shunmugesh and Panneerselvan [4] employed an L27 orthogonal array to design experimental runs for drilling carbon fiber-reinforced plastic (CFRP), using Taguchi and GRA methods for optimization and analysis. Other studies have studied the performance of CO2 laser cutting in processing CFRP sheets [5], laser cut quality of GFRP composites [6] etc. Takahashi et al. [7] analyzed the HAZ and kerf in CFRP composites. The effects of gas flow rate, cutting speed, and laser power on kerf width, HAZ and taper percentage in laser cutting processes was studied by Rao et al. [8]. The consequences of laser drilling on the mechanical and surface properties of GFRP composites was studied by Solati et al. [9]. Despite the extensive research on the optimization of laser drilling processes for different materials, limited studies have focused on kenaf/HDPE composites. Most existing studies have employed traditional single-objective optimization techniques, which may not accurately represent the trade-offs between multiple conflicting objectives. Furthermore, very few studies have utilized hybrid multi-criteria decision-making (MCDM) methodologies to optimize laser drilling processes. The current literature lacks a comprehensive approach that simultaneously optimizes hole quality characteristics, such as ϕ and Ra , in laser-drilled kenaf/HDPE composites using a hybrid MCDM methodology. Additionally, the effectiveness of the CRITIC method for weight allocation in MCDM processes has not been extensively explored in the context of laser drilling optimization of composites. This paper addresses the identified research gap by proposing a novel hybrid CRITIC-MABAC methodology for optimizing the laser drilling process of kenaf/HDPE composites. The key contributions of this paper are— • The development of a hybrid CRITIC-MABAC methodology for optimizing the laser drilling process of kenaf/HDPE composites to achieve minimum ϕ and Ra . • A comparative analysis of the proposed CRITIC-MABAC methodology with the Entropy-MABAC approach, highlighting the effectiveness of the CRITIC method. • The identification of the optimal parametric combination for laser drilling of kenaf/HDPE composites.

106

S. Prabhukumar et al.

2 Problem Description The primary objective of this paper is to examine the hole quality features in laserdrilled kenaf/HDPE composites using a hybrid MCDM methodology. Hybrid CRITICMABAC methodology is used to determine the optimal parametric combination. The idea is to optimize the process parameters to achieve minimum ϕ and Ra . The Laser power and cutting speed are considered as the process parameters. A Kenaf/HDPE composite is fabricated using microwave-assisted compression molding is considered as the workpiece. The experimental dataset is designed using central composite design. More details of the experimentation and the decision matrix for the MCDM process can be found in Tewari et al. [10].

3 Methodology 3.1 CRITIC Method When decision-makers lack the ability to compare multiple standards or hold divergent opinions regarding the relative relevance of distinct criteria under consideration [11], this method is used to estimate impartial criteria weights. Standard deviation and correlation with other evaluation criteria are used to estimate the weight of a given criterion in this method. The jth criterion’s weight is calculated as, Cj wj = n

j=1 Ci

(1)

Cj can be calculated as, Cj = σj

m    1 − cij

(2)

i=1

Here σj is the standard deviation of jth criterion, and cij is the correlation coefficient between ith and jth criteria. 3.2 MABAC Method The University of Defence in Belgrade created a system for making multi-criteria decisions called Multi-Attributive Border Approximation Area Comparison (MABAC) [12]. Alternatives can be ranked and evaluated using MABAC’s predetermined criteria. The procedure accounts for the relative relevance of each criterion. The pseudo code for the MABAC method is as follows— 1. Develop the decision matrix (X ) with m alternatives and n criteria. xij are the elements of X .

Optimizing Laser Drilling of Kenaf/HDPE Composites

107

2. Normalize X to form the normalized matrix R (with elements rij ) using the following rules rij =

xij − xj− xj+ − xj−

rij =

for benefit criteria

xij − xj+ xj− − xj+

for cost criteria

xj+ and xj− are the maximum and minimum values of the J th criterion. 3. Compute the weighted normalized decision matrix V (with elements vij )   vij = wj · rij + 1

(3)

(4)

(5)

wj is the weight of the J th criterion. 4. Compute the border approximation area (BAA) matrix B (with elements bj ) bj =

m i=1

vij

1/m

(6)

5. Compute the distance matrix of alternatives (Q) from the BAA. qij are the elements of Q. Q =V −B 6. Compute the criteria function (Si ) values and ranking the alternatives: n qij , j = 1, 2, . . . , n, i = 1, 2, . . . , m Si = j=1

(7)

(8)

7. Rank the alternatives in descending order of Si values.

4 Results & Discussion 4.1 Optimization with CRITIC-MABAC The experimental data from Tewari et al. [10] is used as the decision matrix in this paper. To find the effect of the weight allocation method two different weight allocation strategies namely the entropy method and the CRITIC method are used. The weights allocated by Entropy method is 11.27% and 88.73% respectively for Ra (C1) and ϕ (C2) respectively. Similarly, the weights allocated by CRITIC method for C1 and C2 is 49.73% and 50.27% respectively. Thus, it is observed that the entropy method produces skewed weights, whereas the CRITIC method produces weights that are almost equalvalued. Table 1 shows the decision matrix and the normalized weighted matrix for both the Entropy-MABAC method and the CRITIC-MABAC method. Figure 1 shows the variation of values with respect to the various experiments. It is observed that the Q-values extend in both the positive and negative directions. For both the weight allocation methods, experiment number 4 is found to have the largest Q-value, indicating that this is the most optimal parametric combination. Experiment number 10 is observed to be the worst parametric combination as per the MABAC method.

108

S. Prabhukumar et al. Table 1. Decision matrix and normalized weighted matrix values

Experimental

Normalized entropy weighted Normalized CRITIC Weighted

C1

C2

C1

C2

C1

C2

5.25

0.314

0.1782

1.5103

0.7863

0.8556

6.61

0.69

0.1330

1.1250

0.5868

0.6373

5.52

0.672

0.1692

1.1434

0.7467

0.6478

3.83

0.056

0.2254

1.7746

0.9947

1.0053

5.25

0.68

0.1782

1.1353

0.7863

0.6431

5.25

0.68

0.1782

1.1353

0.7863

0.6431

6.83

0.736

0.1257

1.0779

0.5545

0.6106

7.22

0.744

0.1127

1.0697

0.4973

0.6060

5.25

0.68

0.1782

1.1353

0.7863

0.6431

6.53

0.922

0.1356

0.8873

0.5986

0.5027

6.12

0.28

0.1493

1.5451

0.6587

0.8753

5.25

0.68

0.1782

1.1353

0.7863

0.6431

5.25

0.68

0.1782

1.1353

0.7863

0.6431

Fig. 1. Variation of MABAC Q-values with respect to experiments

4.2 Parametric Optimization In this section, the effect of the process parameters on the responses is observed. Figure 2 shows the effect of the power on the Ra and the ϕ. It is observed that with an increase

Optimizing Laser Drilling of Kenaf/HDPE Composites

109

in power, the Ra becomes worse. However, in the case of the ϕ, the increase in power increases the ϕ up to a certain power value beyond which the ϕ begins to drop down.

Fig. 2. Effect of power on aggregated values of (a) Ra (b) ϕ

In the context of this study, the MABAC Q-values can be used as a substitute for the combined metric representing the lower power and lower ϕ. Figure 3 shows the effect of power on the aggregated Q-values of the MABAC method. In the case of both the Entropy-MABAC method and the CRITIC-MABAC method the optimal value of power is observed to be 120W.

Fig. 3. Effect of power on aggregated Q-values of (a) Entropy-MABAC (b) CRITIC-MABAC

Figure 4 shows the effect of speed on the Ra and the ϕ. It is noted that an increase in speed gives lower average Ra and lower average ϕ. Figure 5 shows the effect of the speed on the Q-values of the Entropy-MABAC and CRITIC-MABAC methods. It is observed from Fig. 5 that the speed of 4 mm/s is optimal for minimizing the Ra and the ϕ simultaneously.

110

S. Prabhukumar et al.

Fig. 4. Effect of speed on aggregated values of (a) Ra (b) ϕ

Fig. 5. Effect of speed on aggregated Q-values of (a) Entropy-MABAC (b) CRITIC-MABAC

5 Conclusion In this paper, a novel hybrid CRITIC-MABAC method is proposed for optimizing the laser drilling process parameters of kenaf/HDPE composites. The objective of this study is to minimize the ϕ and Ra . The laser power and cutting speed are considered as the process parameters. An experimental dataset, designed in central composite design is used in the case study. The analysis of the case study using the CRITIC-MABAC method reveals that a laser power of 120W and cutting speed of 4 mm/s is the most optimal. Entropy-MABAC approach is also used to compare and validate the findings of CRITIC-MABAC. Thus, the CRITIC-MABAC method is found to be effective in optimizing parametric combination for laser drilling of kenaf/HDPE composites. The findings of this study provide valuable insights for researchers and practitioners working with kenaf/HDPE composites, as the proposed CRITIC-MABAC methodology can be effectively applied to optimize the laser drilling process for achieving superior hole quality characteristics. In addition, the study highlights the importance of the CRITIC method for weight allocation in MCDM processes, as it produces more balanced

Optimizing Laser Drilling of Kenaf/HDPE Composites

111

weight allocations compared to the Entropy method. Future research could explore the application of the CRITIC-MABAC methodology to other composite materials and manufacturing processes, as well as compare its performance with other MCDM methods for process optimization.

References 1. Singh, S., Yaragatti, N., Doddamani, M., Powar, S., Zafar, S.: Drilling parameter optimization of cenosphere/HDPE syntactic foam using CO2 laser. J. Manuf. Process. 80, 28–42 (2022) 2. Solati, A., Hamedi, M., Safarabadi, M.: Combined GA-ANN approach for prediction of HAZ and bearing strength in laser drilling of GFRP composite. Opt. Laser Technol. 113, 104–115 (2019) 3. Palanikumar, K., Latha, B., Senthilkumar, V.S., Davim, J.P.: Analysis on drilling of glass fiber-reinforced polymer (GFRP) composites using grey relational analysis. Mater. Manuf. Process. 27, 297–305 (2012) 4. Shunmugesh, K., Panneerselvam, K.: Optimization of process parameters in micro-drilling of carbon fiber reinforced polymer (CFRP) using Taguchi and grey relational analysis. Polym. Polym. Compos. 24, 499–506 (2016) 5. Riveiro, A., Quintero, F., Lusquiños, F., et al.: Experimental study on the CO2 laser cutting of carbon fiber reinforced plastic composite. Compos. Part A Appl. Sci. Manuf. 43, 1400–1409 (2012) 6. Choudhury, I.A., Chuan, P.C.: Experimental evaluation of laser cut quality of glass fibre reinforced plastic composite. Opt. Lasers Eng. 51, 1125–1132 (2013) 7. Takahashi, K., Tsukamoto, M., Masuno, S., et al.: Influence of laser scanning conditions on CFRP processing with a pulsed fiber laser. J. Mater. Process. Technol. 222, 110–121 (2015) 8. Rao, S., Sethi, A., Das, A.K., et al.: Fiber laser cutting of CFRP composites and process optimization through response surface methodology. Mater. Manuf. Process. 32, 1612–1621 (2017) 9. Solati, A., Hamedi, M., Safarabadi, M.: Comprehensive investigation of surface quality and mechanical properties in CO2 laser drilling of GFRP composites. Int. J. Adv. Manuf. Technol. 102(1–4), 1–18 (2018) 10. Tewari, R., Singh, M.K., Zafar, S., Powar, S.: Parametric optimization of laser drilling of microwave-processed Kenaf/HDPE composite. Polym. Polym. Compos. 29(3), 176–187 (2021) 11. Diakoulaki, D., Mavrotas, G., Papayannakis, L.: Determining objective weights in multiple criteria problems: the CRITIC method. Comput. Oper. Res. 22(1), 763–770 (1995) ´ 12. Pamuˇcar, D., Cirovi´ c, G.: The selection of transport and handling resources in logistics centers using multi-attributive border approximation area comparison (MABAC). Expert Syst. Appl. 42(6), 3016–3028 (2015)

Education, Healthcare, Industry, and Advanced Engineering

Determination of the Optimal Speed of Movement of the Conveyor Belt of the Prototype Weighing Belt Batcher Denis Shilin, Dmitry Shestov, Alexey Vasiliev(B) , and Valery Moskvin National Research University “MPEI”, Moscow, Russian Federation [email protected], [email protected]

Abstract. Today, the agro-industrial complex of Russia is increasing production volumes due to the introduction of high-performance technological lines, and in the field of mixture formation, the processes of production of dry combined products in continuous installations are becoming widespread. Also, the Russian Federation is a large center of light and heavy industry, each of which, at one level or another, involves continuous weighing of components for the production or packaging of bulk mixtures. The weighing accuracy in these units directly affects the cost price and product quality. As a rule, the developers of continuous dosing systems declare dosing accuracy in nominal modes, although in real conditions such systems are characterized by an uneven flow of bulk material, which in turn significantly affects the final dosing error. In connection with the above, the authors of the work developed a prototype of a weighing belt batcher (WBB) and conducted research to determine the optimal range of speeds of the conveyor belt, at which the minimum relative error is achieved. Keywords: Weighing belt batcher · Weighing error · Electric drive · Conveyor · Intelligent algorithm · Fuzzy controller · Weighing mode

1 Introduction The Russian Federation is a large center of light and heavy industry enterprises, each of which, at one level or another of the technological cycle, uses systems for continuous weighing of bulk materials in the manufacture or packaging of various mixtures. At the same time, the cost price and quality of the prepared mixtures are influenced by such indicators as: the productivity of such systems, the accuracy and energy consumption of the weighing process. Typically, such complexes consist of weighing belt batcher, and control systems, the execution of which is implemented both at the hardware and software levels. The main problems of such complexes and the methods of dosing bulk materials used in them are associated with deviations from the required proportions in the preparation of bulk mixtures. In existing technological lines, weighing belt batcher with a dosing error of more than 2.5% are still used. There are hundreds of designs of weighing belt batcher, in which different methods of weight continuous dosing of bulk © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 115–124, 2024. https://doi.org/10.1007/978-3-031-50158-6_12

116

D. Shilin et al.

Readings of strain gauges and , [kg]

mixtures are implemented [11–13]. Typical schemes of weighing belt batcher usually differ in the place of attachment of the strain gauge [11, 12], while the analysis showed that there are no systems on the world market with attachment of sensors from the side of loading materials. The authors of the work have developed an intelligent hardware and software complex for continuous weighing of bulk materials, which consists of a weighing belt batcher with a unique attachment of two strain gauges in the loading area of bulk material and an intelligent algorithm for continuous weighing. The proposed kinematic scheme determines the high scientific significance of the work done due to the absence of such complexes on the world market, in which the kinematic scheme for connecting the strain gauges of the hardware corresponded to the proposed one. The developed intelligent algorithm for continuous weighing is a scientific novelty of the complex. The core of the unique continuous weighing technology is the intelligent fuzzy logic controller [14–18]. The principle of operation of the unique technology for weighing bulk mixtures is based on an automatic system for recording the weight indicators of load cells located on the side of the loaded materials, which, thanks to the use of a fuzzy controller, maintains the speed of the belt conveyor and recalculates telemetry in the developed WBB control cabinet. Figure 1 shows the dependence of the readings of the strain gauges XT and YT on time. Thanks to the developed fuzzy weighing algorithm, the system locates the center of the instantaneous mass of the incoming material along the width H and length L of the conveyor belt, as shown in Fig. 2.

Time, [s] Fig. 1. Time dependence of load cell readings

As seen in Fig. 2, the resulting instantaneous mass is displaced to one side of the conveyor belt. The exact location of the center of mass of the obtained instantaneous mass of material at one time or another, taking into account the dynamics of the induction motor and a unique accounting algorithm, allows you to record the shipped mass with high accuracy [19]. The control system is based on an industrial logic controller. The intelligent hardware and software complex makes it possible to minimize the weighing error (less than 0.5%) by ensuring the operation of the belt conveyor in the maximum loaded mode, which

Determination of the Optimal Speed of Movement of the Conveyor

117

Fig. 2. Determination of the position of the center of mass of the incoming material on the WBB conveyor belt

avoids the accumulation of errors when idling. As a result of the analysis, it was found that one of the main causes of errors when using continuous weighing batchers is the dynamic effects on strain gauges. In this regard, it is necessary to conduct an experiment and empirically determine the permissible speed of the dispenser conveyor belt, which ensures the required weighing accuracy.

2 Mathematical Description of the FC—AM System and Production Line Model with an Electric Drive System Consider the generalized functional diagram of the automated electric drive in Fig. 3, where indicated: C—controller; FC—voltage converter; OR—the object of regulation; FS—feedback sensor. A control signal is input to the system, which sets the value of the output coordinate and can be represented in analog, pulse or digital form.

C

FC

ОR

FS Fig. 3. A generalized functional diagram of an automated electric drive with the control circuit of the output coordinate.

The object of regulation is affected by a disturbing effect Uim in the form of a change: frequency, or voltage of the primary power source of the drive, or static load

118

D. Shilin et al.

moment. The converter can be of any type: electromechanical or static on semiconductor switches. The object of regulation often includes the engine together with the actuator of the actuator, therefore it can be described by various transfer functions. An analog or frequency speed or angle sensor with a gear ratio can be used as a feedback sensor kFS . The difference between the signals given by the output coordinate of the drive and proportional to the actual value enters the regulator, which, in accordance with its transfer function, generates a control signal for the voltage converter UFC , which in turn controls the motor Um to reduce this difference to zero or, if possible, reduce it [20]. UFC = kC (U0 − UFS ),

(1)

UFS = kFS Uout .

(2)

In order to investigate such a system, it is necessary to have a system of differential equations characterizing the dependence of coordinates on external influences and from each other. In general, these equations may be non-linear, i.e. their coefficients may depend on time or coordinate values. Some coordinates of the drive can be a function of the product of external influences and variables, as in our case with a conveyor dispenser [21]. Based on the above generalized functional scheme, the structure of the mathematical model of the technological line of mixing preparation has been developed (Fig. 4).

Speed feedback Conveyor belt speed The task for the speed of the conveyor belt

Controller

Feedback on the volume of material

Frequency converter

МOTOR

Bulk material supply

Conveyor dosing unit with material weight sensors

Fig. 4. The structure of the production line model.

Calculator of mass and volume of material

Determination of the Optimal Speed of Movement of the Conveyor

119

3 Prototype Test Bench Within the framework of this work, tests were carried out of the experimental WBB shown in Fig. 5. The stand includes a WBB (1), consisting of a belt conveyor (2), on the shaft of which an asynchronous motor (3) is installed, and a control system (4) connected with local (5) and remote (6) control systems. The main structural elements of the WBB are: a metal body and a belt conveyor design. These systems were discussed in detail in articles by Shilin et al. [22, 23].

5 3 2

1 4

6

Fig. 5. External view of the WBB pilot plant

WBBs include loading bulk material from a feeder onto a conveyor belt, dynamic weighing along its entire length, and unloading bulk material at the end of its stroke [24–27]. The control system contains the necessary functionality for testing to determine the optimal speed of the conveyor belt, namely: adjusting the speed of the conveyor belt, fixing the shipped mass and other technological settings [28]. The technical characteristics of the device are shown in Table 1.

4 Test Procedure Experimental studies were carried out for various types of bulk materials, the density of which varied from 800 kg/m3 to 2500 kg/m3 . The average duration of the experiment was calculated as the arithmetic mean of the duration of the 3 tests performed. Mass flow rate QM was calculated using the following formula: QM =

me . tm

(3)

120

D. Shilin et al. Table 1. Technical characteristics of WBB

№

Name of metrological and technical characteristics

Value

1

Maximum performance limit (MPL), [tons/hour]

Up to 100

2

Minimum performance limit, [MPL %]

10

3

Limits of permissible error, [MPL %]

0.5

4

Power consumption, not more than, [kW]

1.5

5

Maximum bulk density of material, [tons/m3 ]

1.4

6

Maximum conveyor belt speed, not more than, [m/s]

1.0

7

Conveyor belt width, [m]

0.5

8

Conveyor belt length, [m]

1.0

9

Power supply parameters voltage [V], frequency [Hz]

350 10 [V], 50 1 [Hz]

10

Probability of no-failure operation after 2000 h, [%]

0.92

11

Operating temperature range, [°C]

From minus 10 to plus 40

12

Total average service life of flow meter-batcher, [years]

10

13

Communication interface with the upper level automatic control system

PROFIBUS DP, ETHERNET (ModBus TCP), ModBus RTU, 4–20 mA

14

Degree of protection

IP54

where tm is the average duration of the experiment, s; me is the reference weight of the cargo, kg. The reference mass of the material to be weighed was preliminarily recorded on a floor scale. The volumetric flow rate QV is calculated using the formula: QV =

QM . ρM

(4)

where QM is the calculated mass flow rate, kg/s; ρM —tabular value of material density, kg/m3 . The absolute measurement error  is calculated as the difference between the reference and received masses md :  = me − md .

(5)

The relative measurement error δ is calculated by the formula: δ=

 · 100% . me

(6)

The average value of the measurement error δm was obtained using the following formula:  n   |δi | /N . δm = (7) i=1

Determination of the Optimal Speed of Movement of the Conveyor

121

where δi is the relative error of the i-th test for one measurement, %; N is the number of tests, pcs. Experimental studies were carried out with the fulfillment of the following conditions: (1) the influence of the feeding device on the device body is completely excluded, and, therefore, on the readings of strain gauges; (2) the supply of the weighed bulk material is carried out strictly on the weighing axis (the axis passing through the centers of two strain gauges); (3) the supply of bulk material is carried out uniformly on the entire weighing axis (over the entire width of the flow area of the feeding device). The experiment program includes the following steps: (1) (2) (3) (4)

control weighing of material using a reference balance; setting the speed of the conveyor belt; calibration of strain gauges; reset of the previously recorded readings (the accumulation of the measured mass occurs due to uncorrected sensors); (5) loading the material to be weighed into the feeding device; (6) start of material supply to the conveyor belt and start of timing; (7) when the feeding device is completely empty and the conveyor belt is released, recording the readings of the measured weight and weighing time. After completing all the steps of the experiment, the required number of times, the calculation of the absolute and relative measurement error is carried out according to formulas (5) and (6), respectively. The study of WBB was carried out on the basis of the computer program «Program for control and weighing of bulk material» [29].

5 Determination of the Optimal Speed Range of Movement of the WBB Conveyor Belt As a result of the data obtained, the dependence of the relative measurement error δ on the speed of the conveyor belt ν at different performance of the WBB was built (see Fig. 6). These characteristics were obtained as a result of developing an intelligent control algorithm based on a fuzzy controller. According to the presented data, it can be seen that with a significant increase in the speed of the conveyor belt ν, with different productivity of the WBB, the relative weighing error δ [30, 31]. It was also found that there is an optimal speed range of movement of the conveyor belt ν = 0.25 ÷ 0.75 m/s, at which a relative weighing error is achieved that does not exceed δ = ±0.5%. According to the data obtained, the dependence of the average relative measurement error δm on the performance of the WBB QM (see Fig. 7) was built, obtained during the development of an intelligent control algorithm based on a fuzzy controller in a given speed range of the conveyor belt movement. The characteristic shows that with an increase in the productivity of the WBB QM , the average relative weighing error δm decreases. This is due to the fact that at a low load of strain gauge sensors, their noise significantly affects the resulting indicators of the

D. Shilin et al.

Relative measurement error , [%]

122

Conveyor belt speed , [m/s] Fig. 6. Dependence of the relative measurement error δ on the speed of movement of the conveyor belt ν at different productivity of the WBB

Average relative error

, [%]

mass flow rate of the installation. Accordingly, with increasing productivity, the effect of noise on the mass flow indication is not significant, which leads to a decrease in the relative weighing error.

Mass flow of the weighing tape batcher

, [t/h]

Fig. 7. Dependence of the average relative measurement error δm on the mass flow rate QM WBB

6 Conclusions 1. It was found that there is an optimal speed range of movement of the conveyor belt ν = 0.25 ÷ 0.75 m/s, at which a relative weighing error is achieved that does not exceed δ = ±0, 5% (according to GOST 30124-94 “Scales and weight dispensers of continuous action”) over the entire possible range of productivity QM = 5.0 ÷ 50.0 tons/hour.

Determination of the Optimal Speed of Movement of the Conveyor

123

2. It has been established that the relative weighing error decreases with an increase in the WBB productivity. This is due to the fact that at a low load of strain gauge sensors, their noise significantly affects the resulting indicators of mass flow. Accordingly, as the load (capacity) perceived by the sensors increases, the effect of noise on the mass flow reading is not significant, which leads to a decrease in the relative weighing error. 3. The experiments carried out on the prototype WBB at various intensities helped to determine the optimal speed range of the units and the margin of error, which will allow us to determine the calibration table of the technical characteristics of the WBB line of various productivity and production intensity. 4. WBB can be recommended for mass production for mass introduction in such industries as the production of polymers, household chemicals, pharmaceuticals and medical preparations, the production of rubbers, precious metals, chemical reagents.

References 1. Alspaugh, M.: Bulk Material Handling by Conveyor Belt, Preface, Littleton, 98 p (2004) 2. Boyd, D.C., Panigrahi, S.: Methods and practices of pressure measurements in Silos. In: Design and Selection of Bulk Material Handling Equipment and Systems: Mining, Mineral Processing, Port, Plant and Excavation Engineering, issue 2, pp. 307–335 (2012) 3. Faber, T.E., Lumley, J.L.: Fluid Dynamics for Physicists. Cambridge University Press, Cambridge, 440 p (1995) 4. Vislov, I.S.: A batch feeder for inhomogeneous bulk materials. In: IOP Conference Series: Materials Science and Engineering (2016) 5. Russell, K.: The Principles of Dairy Farming. Farming Press Books and Videos, Ipswich, 370 p (1991) 6. Jackson, W.: Livestock Farming (2004) 7. Lodewijks, G.: The two-dimensional behaviour of belt conveyors. In: Proceedings of the Beltcon 8 Conference, Pretoria, South Africa, pp. 24–26 (1995) 8. Phillips, C.J.C.: Grazing Management and Systems, pp. 188–200 (2015) 9. Shi, F., Minzu, Z., Xiaofeng, S.: The application of iterative learning control algorithm in weighing batch-ing system. In: Mechanical and Electrical Technology, pp. 47–50 (2014) 10. Hua, Z., Zhijiong, L., Rixing, C.: New automatic rubber mixer batching and weighing control system. J. Weighing Apparatus 1–6 (2013) 11. Vidineev, Yu.D., Yanbukhtin, I.R.: Automatic Continuous Dosing of Bulk Materials. Energy, Moscow, 120 p (1974) 12. Pershina, S.V., Katalymov, A.V., Odnolko, V.G., Pershin, V.F., Glinkina, T.M.: Weight Batching of Granular Materials. Mechanical Engineering, Moscow, 260 p (2009) 13. Shestov, D.A.: Analysis of the processes of dosing and weighing bulk materials. In: Proceedings of the International Scientific and Technical Conference Energy Supply and Energy Conservation in Agriculture. GNU VIESH, vol. 3, pp. 109–115 (2012) 14. Karande, A.M., Kalbande, D.R.: Weight assignment algorithms for designing fully connected neural network. IJISA 6, 68–76 (2018). https://doi.org/10.5815/ijisa.2018.06.08 15. Dharmajee Rao, D.T.V., Ramana, K.V.: Winograd’s inequality: effectiveness for efficient training of deep neural networks. IJISA 6, 49–58 (2018). https://doi.org/10.5815/ijisa.2018. 06.06

124

D. Shilin et al.

16. Hu, Z., Tereykovskiy, I.A., Tereykovska, L.O., Pogorelov, V.V.: Determination of structural parameters of multilayer perceptron designed to estimate parameters of technical systems. IJISA 10, 57–62 (2017). https://doi.org/10.5815/ijisa.2017.10.07 17. Awadalla, M.H.A.: Spiking neural network and bull genetic algorithm for active vibration control. IJISA 10(20), 17–26 (2018). https://doi.org/10.5815/ijisa.2018.02.02 18. Abuljadayel, A., Wedyan, F.: An approach for the generation of higher order mutants using genetic algorithms. IJISA 10(1), 34–35 (2018). https://doi.org/10.5815/ijisa.2018.01.05 19. Shilin, D.V., Shestov, D.A., Ganin, E.: Improving the accuracy of weighing bulk materials in a dispenser on-stream flow meter with two strain gauges. Vestnik MEI 3(3), 116–123 (2019). https://doi.org/10.24160/1993-6982-2019-3-116-1233 20. Kang, I.-J., Kwon, J.H., Moon, S.M., Hong, D.: A Control System Using Butterworth Filter for Loss-in-Weight Feeders (2014). https://doi.org/10.7736/KSPE.2014.31.10.905 21. Siva Vardhan, D.S.V., Shivraj Narayan, Y.: Development of an Automatic Monitoring and Control System for the Objects on the Conveyor Belt (2015). https://doi.org/10.1109/MAMI. 2015.7456594 22. Shilin, D., Shestov, D., Ganin, P., Novikov, A., Moskvin, V.: Development of an experimental model of a flow meter-batcher of various intensities. In: Annals of DAAAM and Proceedings of the International DAAAM Symposium, vol. 31, issue 1, pp. 98–103 (2020). https://doi. org/10.2507/31st.daaam.proceedings.013 23. Vasilyev, A.N., Vasilyev, A.A., Shestov, D.A., Shilin, D.V.: Mathematical Modeling of the Work of the Flow-Meter Flowmeter-Doser, pp. 293–299 (2019). https://doi.org/10.1007/9783-030-00979-3_30 24. Kuo, B.C.: Automatic Control Systems, 7th edn. Prentice-Hall Inc, Englewood Cliffs New Jersey, 1003 p (1995) 25. Zhi-yong, Z., Shi-ming, Y., ShiJin, P.: Three-Phase Asynchronous Motor Based on Fuzzy PI Control of Simulink Modeling and Simulation, Mechanical and Electrical Engineering, pp. 53–57 (2012) 26. Rivkin, M.: Bulk Material Handling: Practical Guidance for Mechanical Engineers. Partridge Publishing Singapore, Singapore (2018) 27. Ross, S.M. Simulation. Taylor & Francis Group, Amsterdam, 297 p (2006) 28. Lieberwirth, H.: Design of belt conveyors with horizontal curves. In: Bulk Solids Handling, № # 14, pp. 283–285 (1994) 29. Shestov, D.A., Shilin, D.V., Ganin, P.E.: Computer program 2018661990 Russian Federation, Program for Control and Weighing of Bulk Materials, Applicant and patentee PRIZMA LLC—App. 08/27/2018; publ. 25.09.2018 30. McGlinchey, D.: Bulk Solids Handling. C.O.S. Printers Pte Ltd, Singapore, 290 p (2008) 31. Chaturvedi, D.K.: Modeling and Simulation of Systems Using MATLAB and Simulink. Taylor & Francis, Berkeley, 734 p (2009)

Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio City, Philippines from 2015 to 2018 Guinness G. Maza(B) , Kendrick Jules G. Zante, Clarence Kyle L. Pagunsan, Angela Ronice A. Doctolero, Rostum Paolo B. Alanas, Criselda P. Libatique, and Rizavel C. Addawe Department of Mathematics and Computer Science, College of Science, University of the Philippines Baguio, 2600 Baguio City, Philippines [email protected]

Abstract. This study analyzes the spatial autocorrelation of Acute Bloody Diarrhea (ABD) prevalence in Baguio City, Philippines using records of ABD cases from 2015 to 2018. The Global Moran’s I was used to identify the spatial pattern, while the Local Moran’s I was utilized to determine if hotspots exist within the area. Results showed a positive spatial autocorrelation in 2015 at a 0.05 significance level, indicating a clustered pattern of ABD prevalence. For the years 2016 to 2018, no significant spatial patterns were found. ABD hotspots were identified for each year. Clustering was observed in the central parts of Baguio City. Keywords: Spatial analysis · Spatial autocorrelation · Hotspots · Acute bloody diarrhea · Moran’s I

1 Introduction Acute Bloody Diarrhea (ABD) is characterized by abdominal pain, dehydration, and consecutive bowel movements with a loose, watery stool containing blood. ABD is also known as dysentery and can be caused by various enteric pathogens and noninfectious gastrointestinal illnesses [1]. It has been linked to Escherichia coli and other Shiga-toxinproducing E. coli, a group of bacteria found in the intestines of humans and animals that can cause symptoms through contaminated meat or water consumption [2, 3]. Diarrhea-related diseases have become a worldwide health concern, with nearly 1.7 billion cases and are responsible for the deaths of around 525,000 children annually [4]. The number of reported food and waterborne diseases (FWBDs) in Baguio City significantly increased from 410 in 2005 to 1,938 in 2010, among which ABD had the highest incidence rate [5]. Concerns of diarrhea epidemic remain as Baguio City continues to face problems of water supply and access to potable water [6]. Environmental factors can alter the exposure pathways of FWBDs and can be influenced by climatic conditions, hence impacting the transmission and reproduction of pathogens [7]. Therefore, understanding and addressing the burden of ABD and other © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 125–133, 2024. https://doi.org/10.1007/978-3-031-50158-6_13

126

G. G. Maza et al.

FWBDs require the study of spatial patterns of disease occurrence. Spatial autocorrelation, defined as the statistical relationship between the values of a variable at different locations, can provide insights into the factors contributing to disease patterns and help identify potential interventions for time- and cost-efficient disease control [8–10]. This study aims to (i) examine the spatial patterns of ABD among the barangays— the smallest political unit of the Philippines—of Baguio City, (ii) identify the hotspots of ABD from 2015 to 2018, and (iii) determine whether the identified hotspots demonstrate any significant patterns in the spread of the disease. The results of this study will be used to assist the local government and health authorities in implementing policies and programs aimed at preventing and controlling the spread of ABD in Baguio City and similar communities, aligning with the United Nations Sustainable Development Goal 3: Good Health and Well-being. The rest of this study is organized as follows. The methodology, area of the study, data collection, the Moran’s I index, and the research instruments are discussed in Sect. 2. The results of the descriptive and inferential statistics are presented in Sect. 3. Lastly, the conclusion and recommendations of this study are presented in Sect. 4.

2 Methodology This section presents all the details regarding the methods applied in the collection, computation, and analysis of the data. This includes the scope, formulas, and instruments used in the study. 2.1 Area of the Study The study area comprises the whole of Baguio City. It is composed of 20 administrative districts and 129 barangays. The City is situated in the province of Benguet, surrounded by different municipalities such as La Trinidad on the North, Itogon on the East, and Tuba on the South and West borders [11]. 2.2 Data Collection The Exploratory Data Analysis (EDA) team of the University of the Philippines Baguio (UPB) and the Baguio City Health Services Office (BCHSO) provided the records of ABD cases in Baguio City from 2015 to 2018. The shapefiles (SHP files) containing the geographical features of the barangays of Baguio City were also provided by the EDA team of UPB. The factors considered for the study include the stool culture results— both positive and unknown, the age of the patient in years, and the barangay where the patients reside. The ABD prevalence was also calculated by dividing the total number of cases per barangay by the total population per barangay in the given year.

Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio …

127

2.3 Moran’s I Index The Global Moran’s I Index, an indicator for global spatial autocorrelation [12], was utilized to assess if there exists a spatial correlation of ABD prevalence among the barangays of Baguio City. A positive value of I would suggest a positive correlation or clustered pattern, while a negative value indicates a negative correlation or dispersed pattern. The null hypothesis for this study is I = 0, which implies complete spatial randomness. The formula for Global Moran’s I is   n ni=1 nj=1 wij zi zj (1) I=  S0 ni=1 zi2   where S0 = ni=1 nj=1 wij and zi = xi − x¯ . The variable xi is the ABD prevalence in each barangay i, xj is the ABD prevalence of the neighbors of i, n is the total number of barangays in the study, x¯ is the mean ABD prevalence of all barangays, and wij is the measure of the weight of barangay i and j that demonstrates the sharing of a boundary between the two. This study utilized the Rook contiguity weights matrix, which assigns a value of 1 to neighboring locations, where  1, if location i and j share a common borderline, and i = j; wij = (2) 0, otherwise. The weights wij were then divided by the number of neighbors of i. This process is also known as row standardization, which ensures that each neighbor j contributes equally to the value of i. The following formula for the z-score was used to test the significance of Global Moran’s I under randomization. Here, I − E(I ) zI = √ V (I )

(3)

where the expected value E(I) is computed as E(I ) = −

1 . n−1

(4)

The variance V (I) of the Global Moran’s I [13] is computed as     n (n2 − 3n + 3)S1 − nS2 + 3S02 − a (n2 − n)S1 − 2nS2 + 6S02 V (I ) = − E(I )2 (n − 1)(n − 2)(n − 3)S02 (5) 2      where S1 = 21 ni=1 nj=1 (wij + wji )2 , S2 = ni=1 ni=1 wij + nj=1 wji , and a = n 4 i=1 zi n 2 2 . i=1 zi The resulting values from computing the Global Moran’s I index will only identify the spatial pattern and not provide the specific location of clusters [14]. Thus, the

128

G. G. Maza et al.

Local Moran’s I of the Local Indicator of Spatial Association (LISA) principle was implemented in this study to specify the locations of possible hotspots. In each i, n j=1 wij (xi − x¯ )(xj − x¯ ) n . (6) Ii = 2 i=1 (xi − x¯ ) The Monte-Carlo method was used to test the significance of the Local Moran’s I. It formulates a pseudo-p-value using the following formula, p=

R+1 M +1

(7)

where R is the total number of random simulations greater than or equal to the computed Local Moran’s I, and M is the number of permutations. Pseudo p-values that are less than or equal to 0.05 will be considered hotspots. 2.4 Research Instruments Three free and open-source software were utilized for data visualization and statistical computation. First, RStudio was utilized for data visualization and computing the Global Moran’s I. The specific packages used were sf [15] for importing SHP files, spdep [16] for the Moran’s I analysis, readxl [17] to analyze data inside excel files, and plotrix [18] for specialized plots. Second, Quantum Geographic Information System (QGIS) [19] was used to edit and add variables to the SHP file of Baguio City. QGIS was also utilized to produce the geometric map of the overlapping hotspots. Lastly, GeoDa [20] was used to compute the Local Moran’s I and to generate LISA cluster maps.

3 Results and Discussion This section shows the computed statistics and analysis of the data. This includes graphs, maps, and tables containing the summary of results. 3.1 Descriptive Statistics A total of 948 cases of ABD were reported in Baguio City from 2015 to 2018. Figure 1 shows the age group distribution of ABD patients. The age grouping was derived from the Philippine Health Statistics’ standard, which is also based on the World Health Organization’s standard [21]. Most of the reported cases were children and young adults. The age group 0 to 4 years old has the highest number of cases at 213, while the age group 60 to 64 has the lowest number of cases at 21. According to Wardlaw et al. [22], children are more likely to come into contact with unsanitary environments, especially when unsupervised. They also have weaker immune systems compared to adults, making them more susceptible to ABD compared to older age groups. Figure 2 shows the number of monthly ABD cases from 2015 to 2018. A total of 347 ABD cases were reported in 2015, 260 cases in 2016, 215 cases in 2017, and 126 cases in 2018. Therefore, the number of ABD cases from 2015 to 2018 has decreased.

Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio …

129

Fig. 1. Age distribution of ABD patients from 2015 to 2018.

Fig. 2. Monthly ABD cases from 2015 to 2018.

3.2 Inferential Statistics The results of the Global Moran’s I are shown in Table 1. There is a positive spatial correlation in 2015 with a Global Moran’s I of 0.1145 and a p-value of 0.009, indicating an ABD prevalence in Baguio City with a clustered pattern at 0.05 significance level. The p-values of the Global Moran’s I from 2016 to 2018 are greater than 0.05, implying no significant patterns of ABD prevalence. The Local Moran’s I was utilized to determine the clusters per year from 2015 to 2018, as shown in the LISA cluster maps in Fig. 3. A High-High relationship indicates that barangays with a high ABD prevalence surround a barangay with a high ABD prevalence. A Low-Low relationship suggests that barangays with a low ABD prevalence surround a barangay with a low ABD prevalence. A High-Low relationship indicates that barangays with a low ABD prevalence surround a barangay with a high ABD prevalence. A Low-High relationship suggests that barangays with a high ABD prevalence surround a barangay with a low ABD prevalence. The clustering pattern in 2015 was identified around the central parts of Baguio City, specifically barangays Harrison-Claudio Carantes, Session Road Area, Salud Mitra, and Engineer’s Hill.

130

G. G. Maza et al. Table 1. Global Moran’s I of ABD prevalence in Baguio City

Year

Moran’s I

p-value

2015

0.1145

0.009∗

2016

0.0393

0.1486

2017

−0.0064

0.5104

2018

0.0729

0.0513

∗ Significant at 0.05

Fig. 3. LISA cluster maps of ABD prevalence for (a) 2015, (b) 2016, (c) 2017, and (d) 2018

A total of 16 barangays were identified as ABD hotspots in 2015, 14 in 2016, 16 in 2017, and 19 in 2018 at a 0.05 significance level. The test was done in GeoDa using Monte-Carlo simulation at 999 permutations. Figure 4 shows the overlapping of hotspots with 4-digit labels. That is, barangays highlighted as 1111 were hotspots of ABD from 2015 to 2018, while those labeled as 0000 were not hotspots from 2015 to 2018. Barangay Malcolm Square-Perfecto, located in the center of Baguio City, was a hotspot of ABD from 2015 to 2018 as shown in Fig. 4. Barangays Lourdes Subdivision Proper, Camp Allen, and Bagong Lipunan were all ABD hotspots from 2016 to 2018. Barangays

Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio …

131

identified as ABD hotspots for two or more years were mostly situated at the center of Baguio City.

Fig. 4. Hotspots of ABD from 2015 to 2018 where 1 indicates a hotspot in a particular year; otherwise, 0. For example, 1100 means that the barangay was only a hotspot for 2015 and 2016.

4 Conclusion and Recommendations The descriptive statistics of monthly cases showed decreased ABD cases from 2015 to 2018. Most reported cases belonged to younger age groups, with the highest being children aged 0 to 4 years old. The spatial analysis result suggests a positive spatial correlation in 2015, implying a clustered pattern of ABD, while spatial randomness was found in 2016, 2017, and 2018. The LISA cluster maps showed that the ABD hotspots from 2015 to 2018 were around the central parts of Baguio City, with barangay Malcolm Square-Perfecto as a hotspot for all four years. The researchers recommend that the local government and health authorities look into the health status of the identified ABD hotspots and surrounding areas and assess potential risks of ABD infection among the citizens. This study also recommends further investigation into other potential variables that contribute to ABD incidence in Baguio City, such as environmental factors and the socio-demographic profile of the patients. Acknowledgement. This study is supported by the University of the Philippines Baguio (UPB) through its DATIVA-Rico Foundation Funds and the Department of Mathematics and Computer Science International Publication Awards Funds. The researchers would like to thank the BCHSO and the EDA team of UPB for the data and SHP files used in the study.

132

G. G. Maza et al.

References 1. Holtz, L.R., Neill, M.A., Tarr, P.I.: Acute bloody diarrhea: a medical emergency for patients of all ages. Gastroenterology, vol. 136, pp. 1887–1898. Elsevier (2009). https://doi.org/10. 1053/j.gastro.2009.02.059 2. Talan, D.A., Moran, G.J., Newdow, M., Ong, S., Mower, W.R., Nakase, J.Y., Pinner, R.W., Slutsker, L.: Etiology of bloody diarrhea among patients presenting to United States emergency departments: prevalence of Escherichia coli O157:H7 and other enteropathogens. In: Clinical Infectious Diseases, vol. 32, pp. 573–580. The University of Chicago Press (2001). https://doi.org/10.1086/318718 3. Rane, S.: Street vended food in developing world: hazard analyses. Indian J. Microbiol. 51, 100–106 (Springer) (2011). https://doi.org/10.1007/s12088-011-0154-x 4. World Health Organization: Diarrhoeal Disease (2017). Last Access 9 Jan 2023. https://www. who.int/news-room/fact-sheets/detail/diarrhoeal-disease 5. Padilla, J.R.F., Pilar, K.C.N., Bitanga, C.A.G., Bumengeg, L.N., Addawe, R.C.: Incidence of food and water-borne diseases in Baguio City. In: The 4th Innovation and Analytics Conference and Exhibition, vol. 2318, p. 050024–5 (2019). https://doi.org/10.1063/1.512 1129 6. Mendoza, L.C., Cruz, G.A., Ciencia, A.N. and Penalba, M.A.: Local policy and water access in Baguio City, Philippines. Int. J. Soc. Ecol. Sustain. Dev. (IJSESD) 11(1), 1–13 (2020). https://doi.org/10.4018/IJSESD.2020010101 7. Semenza, J.C., Herbst, S., Rechenburg, A., Suk, J.E., Höser, C., Schreiber, C., Kistemann, T.: Climate change impact assessment of food-and waterborne diseases. In: Critical Reviews in Environmental Science and Technology, vol. 4, pp. 857–890. Taylor & Francis (2012). https://doi.org/10.1080/10643389.2010.534706 8. Hershey, C.L., Doocy, S., Anderson, J., Haskew, C., Spiegel, P., Moss, W.J.: Incidence and risk factors for malaria, pneumonia and diarrhea in children under 5 in UNHCR refugee camps: a retrospective study. Conflict Health 5, 7 (BioMed Central) (2011). https://doi.org/10.1186/ 1752-1505-5-24 9. Kamath, A., Shetty, K., Unnikrishnan, B., Kaushik, S., Rai, S. N.: Prevalence, Patterns, and Predictors of Diarrhea: A Spatial-Temporal Comprehensive Evaluation in India. BMC Public Health, vol. 18, pp. 1–10. Springer, Heidelberg (2018). https://doi.org/10.1186/s12889-0186213-z 10. Su, T., Liu, Y., Zhao, W., Yu, Q., Xie, Y., Li, Q., Qi, S.: Epidemiological characteristics and spatial-temporal cluster analysis of other infectious diarrhea in Hebei Province from 2015 to 2020. Chin. J. Dis. Control Prev. 26(2), 175–181 (2022). https://doi.org/10.16462/j.cnki.zhj bkz.2022.02.009 11. City Environment and Parks Management Office: City Report, Baguio City, Philippines. In: Eighth Regional EST Forum in Asia, p. 1 (2014). Last Access 18 Feb 2023: Retrieved from https://www.uncrd.or.jp/content/documents/21038EST-City-Report_Philippines-Baguio.pdf 12. Moran, P.A.P.: The interpretation of statistical maps. J. R. Stat. Soc.: Ser. B (Methodological) 10, 243–251 (JSTOR) (1948). https://doi.org/10.1111/j.2517-6161.1948.tb00012.x 13. Bivand, R.S., Wong, D.W.S.: Comparing implementations of global and local indicators of spatial association. TEST 27, 716–748 (2018). https://doi.org/10.1007/s11749-018-0599-x 14. Anselin, L.: Local indicators of spatial association-LISA. Geogr. Anal. 27(2), 93–115 (1995). https://doi.org/10.1111/j.1538-4632.1995.tb00338.x 15. Pebesma, E.J.: Simple features for R: standardized support for spatial vector data. R. J. 10(1), 439 (2018). Last Access 24 Dec 2022. Retrieved from https://www.pebesma.staff.ifgi.de/RJw rapper.pdf

Spatial Analysis: Cases of Acute Bloody Diarrhea in Baguio …

133

16. Bivand, R., Altman, M., Anselin, L., Assunção, R., Berke, O., Bernat, A., Blanchet, G.: Package ‘spdep’. The Comprehensive R Archive Network, vol. 604, p. 605 (2015). Last Access 24 Dec 2022. Retrieved from https://www.cran.microsoft.com/snapshot/2017-09-17/ web/packages/spdep/spdep.pdf 17. Wickham, H., Bryan, J., Kalicinski, M., Valery, K., Leitienne, C., Colbert, B., Hoerl, D., Miller, E. and Bryan, M.J.: Package ‘readxl’. Version 1.3, vol. 1 (2019). Last Access 24 Dec 2022. Retrieved from https://www.cran.microsoft.com/snapshot/2019-03-09/web/packages/ readxl/readxl.pdf 18. Lemon, J., Bolker, B., Oom, S., Klein, E., Rowlingson, B., Wickham, H., Tyagi, A., Eterradossi, O., Grothendieck, G., Toews, M. and Kane, J.: Package ‘plotrix’. Vienna: R Development Core Team (2015). Last Access 24 Dec 2022. Retrieved from https://www.mran.revolu tionanalytics.com/snapshot/2020-04-25/web/packages/plotrix/plotrix.pdf 19. QGIS Development Team: Quantum Geographic Information System. Last Access 21 Dec 2022. www.qgis.org 20. Anselin, L., Syabri, I., Younghin, K.: GeoDa: an introduction to spatial data analysis. Geogr. Anal. 38(1), 5–22 (2006). https://doi.org/10.1007/978-3-642-03647-7_5 21. Epidemiology Bureau, Department of Health: The 2019 Philippine Health Statistics, vol. 1, pp. 85–89. Philippine Health Statistics (2019). Last Access 5 Feb 2022. Retrieved from https://www.doh.gov.ph/sites/default/files/publications/2019PHS_Final_092121.pdf 22. Wardlaw, T., Salama, P., Brocklehurst, C., Chopra, M., Mason, E.: Diarrhoea: why children are still dying and what can be done. The Lancet 375(9718), 870–872 (2010). https://doi.org/ 10.1016/s0140-6736(09)61798-0

The Economic Dimensions of the Non-communicable Diseases: A Panel Data Study Sergio Arturo Domínguez-Miranda(B)

and Roman Rodriguez-Aguilar

Facultad de Ciencias Económicas y Empresariales, Universidad Panamericana, Mexico University, Augusto Rodin 498, 03920 Mexico City, Mexico [email protected]

Abstract. Noncommunicable diseases (NCD) are the leading causes of death and the main public health problem worldwide [1] and are associated with an acute picture of dyslipidemia that is part of the main risk factors along with smoking, sedentary activities, incorrect diet, and genetic factors that have caused diseases such as metabolic syndrome, oncological, cardiology, neurological and respiratory diseases. NCD are a major problem in low-income and middle-income countries, consuming increasing proportions of health care budgets. NCD are among the leading causes of disability and ill health and are the leading cause of preventable and premature death, having a significant impact on economies. NCD generate large out-of-pocket health costs for both individuals and families, as well as huge health outlays in national budgets. The analysis proposed can provide critical information to monitor trends in population health outcomes, recognize the pattern of diseases and injuries affecting premature mortality and disability. This paper shows an approach for the identification of the behavior of four main NCD (cardiovascular diseases, respiratory diseases, diabetes mellitus, and neoplasms) along thirteen countries selected and various economic variables related to the health and work issues from 1961 to 2021 with unbalanced data. The primary focus is the analysis on mortality in population within working age. Keywords: Noncommunicable diseases · Econometric models · Panel data · Working age · Health economics

1 Introduction 1.1 Overview of Noncommunicable Diseases Globally and Its Economical Dimensions For just over 20 years, NCD have occupied the first places as causes of general death: heart disease, stroke, and diabetes mellitus, being in the first, second and ninth place respectively [2]. Researchers in conjunction with the Pan American Health Organization [3] estimated that from 2010 to 2030 an expenditure of 47 trillion dollars is expected on the treatment of NCD, which is equivalent to a loss of 48% of annual global GDP © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 134–144, 2024. https://doi.org/10.1007/978-3-031-50158-6_14

The Economic Dimensions of the Non-communicable Diseases

135

and impacts 4% of annual GDP in middle- to low-income countries. Other ones have identified that about 60% of cases reviewed out-of-pocket expenses represent 20% to 30% of their income [4] estimated that in the United States of America there is a loss of productivity in a range of 8539 to 10,175 dollars for cardiovascular diseases and 1,962,314 dollars per year due to diabetes mellitus with a premature mortality of 49% [5]. Considering the effects of prevention and promotion on health, companies are susceptible to the effects generated by NCD in the absence of strategies for employee health, causing presenteeism, poor job performance, and loss of productivity. Several chronic and acute diseases are increasingly important, and the medical industry is changing drastically due to the need for real-time diagnosis and monitoring of long-term health conditions. However, in the absence of data analysis, and information, it is difficult to proceed with any care program, especially in support of the working-age population to prevent and predict the behavior of health indicators in such a way that the main question is: What is the behavior of NCD in the world and its relation to relevant economic variables in the population whose age is within the work period? Some research in the past had tried to explain certain social different variables with different economic models [6, 7], some other developed a classification method using social-economic data for just one year [8], or for validating public policies base on one data base along the time [9], but there is still a gap showing how some variables along the time could affect the one of the main social problems, the health status. This research article proposes to review the quantitative evidence of noncommunicable diseases for selected countries trying to better understand the impact of relevant variables, analyzing the period from 1961 to 2021 through an unbalanced data panel for a working age. The work methodology will be mentioned, considering the way of extracting and filtering variables, the mathematical model used, and the results obtained from the analysis based on panel data in each NCD. The objective is to understand the behavior of economic indicators, lifestyles, and diet in NCD mortality to find patterns of behavior, understand the impact of the most significant variables, and be able to support both companies and policyholders of public decisions to take the pertinent measures to improve health indicators, and therefore have an impact on productive indicators.

2 Methodology Econometrics with economic models can help identify whether there is a statistical problem so that the researcher is able to interpret, summarize, and describe the situation coherently. In the same way, it helps us to make valid inferences for a larger population of individuals with similar characteristics and the samples are analyzed over time in the panel data, usually before the intervention of the program and after the intervention, getting two dimensions: spatial or structural and temporal [10]. It allows to model individual observable differences and comes to solve one of the weaknesses of the classic regression model, which is presence of heteroscedasticity. Being a study focused on the behavior of NCD within the working period, age filters between 20 and 64 years were used, as well as thirteen countries selected non-randomly considering variability in their GDP, being analyzed: South Korea, China, India, Canada,

136

S. A. Domínguez-Miranda and R. Rodriguez-Aguilar

Colombia, Turkey, United States of America, Argentina, Mexico, Chile, Australia, Israel, Brazil, Japan, France, Germany, and Poland in the period from 1961 to 2021 generating a non-balanced panel. Databases and variables shown in Table 1 from IHME [11], OECD [12], WHO [13] and FAO [14] were used. Prevalence, incidence, and mortality of the four main NCD were included [11], being cardiovascular diseases (which includes stroke and ischemic heart disease, among others), respiratory diseases (only includes NCDs such as chronic obstructive disease), diabetes mellitus (which includes the conditions that this disease generates such as kidney diseases) and neoplasm (where all types of tumors are included in the population). Population means of health indicators were selected [12]. Relevant elements involving lifestyles (such as years of life lost or exposure to pollution), economic variables (such as investment and government spending on health) were used and variables related to work (such as productivity or economically active population) using the filters initially considered [13]. Finally, indicators related to eating behavior (such as consumption of meat or sugary drinks) were included [14]. An unbalanced data panel matrix was integrated, obtaining observations from 1961 to 2021 getting a matrix of dimensions 61 × 578. Several methods were used to eliminate multicollinearity to avoid the problem that could lead to statistically insignificant variables and, on the other hand, to the inclusion of new variables. Traditionally, the statistical methods that try to explain an observed phenomenon through a series of variables have been treated by linear regressions, using the OLS method. To evaluate this condition, the Hausman test can be used, first making an estimation by means of OLS and then making a panel of data, to finally run the Hausman analysis [15]. Since there may be differences in the behavior of individuals, the source of sample variation is important in the formulation and estimation of our economic model. For the case where the endogenous variable Y i and several explanatory variables X 1 , X 2 ,…, X k are observed for each individual and period, we could have the model of constant coefficients or data pool, where the intersection and the coefficients are constant with respect to time and between individuals. Therefore, the possible differences between individuals and different moments of time are assimilated to the random term. Using the formula (Eq. 1): Yit = β1 + β2 Xit + β3 Xit + . . . + βk Xit + uit

(1)

where Y is the variable focused on the outcome of NCD, in this case mortality as the main dependent variable, also evaluating prevalence and incidence. The i value was used for the cross-sectional variables related to the countries, and t for the time from the period 1961 to 2021. The difference, between the countries in their different variables and time periods, was added to the model, seeking to reduce autocorrelation problems because the variance of the disturbances can be different with respect to countries or over time, and/or heteroscedasticity. As all the slopes or coefficients of the variables are constant, but not the intersection, the model was adjusted looking for heterogeneity in the behavior of the teaching units

The Economic Dimensions of the Non-communicable Diseases

137

Table 1. Data bases used for analysis Variables

Unit of measure

Incidence

Number of people

[11]

Permanence

Number of people

[11]

Deaths

Number of people

[11]

Average systolic blood pressure

MmHg

[12]

Average HDL cholesterol

Mmol/L

[12]

Prevalence of hypertension

Average years

[12]

Average rapid glucose

Mmol/L

[12]

BMI_Promedio

kg/m2

[12]

Out-of-pocket expenditure

USD per capita

[13]

Government Investment

USD per capita

[13]

Years lost to live

Years per 100,000 habitants

[13]

Government investment

% Of GDP

[13]

Economic active population

% With respect to population

[13]

Productivity

GDP/hour/person

[13]

Health expenditure

USD per capita

[13]

Insurance spending

USD per capita

[13]

Government health expenditure

USD per capita

[13]

Total out of pocket expenditure

USD per capita

[13]

Total population

Millions of people

[13]

Pollution exposure

Micrograms/mts3 /population

[13]

Wheat and products Food supply quantity

Kcal/person/day

[14]

Processed rice—Food supply quantity

Kcal/person/day

[14]

Consumption of alcoholic beverages

Liters per capita

[14]

Beef Food supply quantity

Kcal/person/day

[14]

Meat consumption

Kilograms per capita

[14]

Maize and products Food supply quantity

Kcal/person/day

[14]

Food supply sweeteners

Kcal/person/day

[14]

collected through the independent terms, that is, in the case of equality of the mean values of the explanatory variables across countries, the mean value of the dependent variable would be different. Therefore, various variables (D) relative to the variables from the databases were assigned to model the independent parameter for each of the countries (i) (Eq. 2). Yit = α1 + α2 D2 + α3 D3 + . . . + αN DN + β2 X2it + β3 Xit + uit

(2)

138

S. A. Domínguez-Miranda and R. Rodriguez-Aguilar

Variables were transformed into deviations from their temporal mean for every case. To determine if the heterogeneity of the model comes from the differences in the independent term, and, therefore, that the model can be conducted by means of a test based on the F statistic, which is formulated as follows (Eq. 3): FN −1,NT −k−N =

R2G −R2rest N −1 1−R2G NT −k−N

(3)

where R2G is the coefficient of determination of the general model and R2rest is the coefficient of determination of the restricted model called covariance analysis. Considering that the coefficients are constant, but the intersection varies according to the countries and time, the model can be used (Eq. 4): Yit = α1 + α2 D2 + α3 D3 + . . . + αN DN + γ1 + γ2 + . . . γt + β2 X2it + β3 Xit + uit (4) For the multivariate case: Yit = β1 + βk Xik + uit where Y is a matrix of k explanatory variables derived from the one reflected in the different databases for each one of the countries so that: y1t y2t where is the Yit vector containing the information of the countries (i) in Yit = : y Nt all t of selected time. X kit is the matrix of observations of the explanatory variables of the indicators used according to Table 1 (k), for the countries (i), from 1961 to 2021 (t). X11t X21t β1 u1t Xk1t .. X X22t Xk2t β u in turn βk = 2 and Uk = 2t vector containing the Xkit = 12t .. : : : : : .. X 1Nt X 2Nt X kNt βN u Nt t random perturbations of each individual variables. The parameter vector α i = [α1 , α2 ,…, αN ] collects the individual effects. The assumptions made in these models are fundamentally the non-correlation between the perturbations of each of the groups, and the temporal non-correlation, and that the variances of the perturbations are homoscedastic and not self-correlated, meaning that:  E[uit ] = 0; var[uit ] = σ 2 ; cov uit , ujs = 0. The mortality of the four main NCD (Cardiovascular, diabetes mellitus, neoplastic, respiratory) was selected a dependent variable (Y ), in addition, it was verified the variables of the state of health and diet that could have a greater impact on the disease were analyzed based on the literature regarding the behavior of the disease in all the variables those related to health expenditure, social determinants and population growth (k). Subsequently, from the variables used, and using the pooled least squares in E-views software application, those that were not significant were eliminated, leaving only those that generated statistical significance. Three models were generated for mortality derived from cardiovascular diseases, three derived from diabetes mellitus, one for neoplastic diseases and two for respiratory

The Economic Dimensions of the Non-communicable Diseases

139

diseases due to different explanations found variating indicators within the models. In addition to explain more information with the Neoplasm disease, the variable of incidence was used due to the high impact in the main variable and explained later. Among the estimated models, the most appropriate ones were selected based on the overall significance, the number of explanatory variables and R2 .

3 Results 3.1 Cardiovascular Table 2 shows in model 1 that incidence (t-value = 35.32) and investment in health (tvalue = −25.06) are the elements with the greatest significance (both with p = 0.000), which reinforces the concept if investments in health are reduced, the impact will be growth in mortality. On the other hand, incidence is one of the main causes of mortality, which agrees with the literature. Additionally, the working population is observed as relevant findings with a negative impact on the indicator, this probably due to the activity generated by work in people who maintain physical activity. It´s also observed that when using a lag of value 3 (equivalent to 3 years) in the average body mass index, this impacts on cardiovascular disease, this can be understood that the impact of not maintaining a healthy life is not reflected in the impact of the disease until later. Similarly, food serves as a relevant element in cardiovascular diseases as it can be observed with the consumption of sweetening beverages (p = 0.000; t = 9.8) as well as meat consumption (p = 0.000; t = 5.7). 3.2 Diabetes Mellitus It is observed according to Table 3 in model 1 that the consumption of wheat (p = 0.071), rice (p = 0.0 76) and sweetening beverages (p = 0.045) has a relevant impact on the evolution of mortality due to diabetes mellitus, in the case of wheat and rice products it is again shown, that using lags of order 5 and 3 respectively, impacts on mortality, an additional indication that correct feeding is essential. Regarding model 2, population indicators were used, where population growth (p = 0.000) is the most relevant. Additionally, the working population shows a positive significance (p = 0.026), unlike cardiovascular diseases, it is likely that the explanation is that, although work generates physical activation, it does not necessarily go hand in hand with food, a situation that agrees with the literature when showing that diabetes diseases, although it is multifactorial, food is of great relevance. As relevant data for model 3 it is shown in this case that the body mass index shows lag at one year (p = 0.008) is the element with greater significance along with the incidence of this disease. 3.3 Neoplasm Diseases Table 4 shows only one construction model to explain what happens with Neoplasm disease, the variables that explain this phenomenon seem logical at the beginning because they depend on the incidence (p = 0.000) and the total population mainly (p = 0.000),

140

S. A. Domínguez-Miranda and R. Rodriguez-Aguilar Table 2 PLS results—cardiovascular diseases

Dependent variable: Mortality from cardiovascular diseases Independent variables

Model 1

Mean systolic pressure

673.10(0.000)

Prevalence of hypertension

−420.89(0.000)

Body mass index (With 3 lags)

1249.47(0.000)

Investment in health

−3.98(0.000)

Model 2

Model 3

4612.50(0.005) −2232.81(0.084)

Total population

1154.79(0.000)

Working population

−635.31(0.000) 1798.61(0.002)

Out-of-pocket expense

15.93(0.000)

Cardiovascular disease incidence

0.03(0.000)

0.051(0.000)

Cardiovascular disease prevalence

1907.92(0.000)

−0.002(0.000)

Meat consumption (With 5 lags)

128.57(0.000)

Consumption of sweetened beverages 422(0.000) AR(1)

1.62(0.000)

1.48(0.000)

AR(2)

−0.68(0.000)

−0.98(0.000)

AR(3)

0.5(0.000)

AR(8)

−0.028(0.031)

R2

0.9991

0.9992

0.9996

R2 Adjusted

0.9990

0.9992

0.9996

Standard regression error

668.44

2227.98

11668.75

Durbin-Watson

1.81

2.022

2.03

Remarks

73

400

322

so we proceeded to carry out a second approach to understand in detail the incidence as a response to the understanding of mortality. Table 5 shows the application of a semilogarithmic system in the dependent, to understand the behavior of the growth of the incidence in Neoplasm diseases. In this sense, it’s shown that there are elements that impact on this disease derived from food, but the impact is not so high as it has low coefficients compared to diabetes mellitus or cardiovascular disease. In this model, per capita alcohol consumption now appears (p = 0.068; t = 1.8), although with low explanation. 3.4 Respiratory Diseases Table 6 now shows the analysis of respiratory diseases where 2 models are used, in the first shows a logical element that is derived from population growth (p = 0.000), however in a second model there are relevant elements such as exposure to pollution (p = 0.002; t = 3.03) that is relevant in the analysis, Although the incidence also appears

The Economic Dimensions of the Non-communicable Diseases

141

Table 3 PLS results—diabetes mellitus Dependent variable: Mortality from diseases caused by diabetes mellitus Independent variables

Model 1

Wheat consumption (5 lags)

13.01(0.071)

Rice consumption (3 lags)

34.44(0.076)

Consumption of sweetened beverages

87.12(0.047)

Model 2

Model 3

Body mass index (1 lag)

1650(0.008)

Average fasting blood glucose

−7249.77(0.012)

Total population

63.43(0.000)

61.42(0.000)

Working population

65.54(0.026)

Productivity per hour per person

−17.91(0.014)

Incidence of diabetes mellitus diseases

0.014(0.000)

Prevalence of diabetes mellitus diseases

−0.0006(0.004)

AR(1)

1.35(0.000)

1.47(0.000)

1.41(0.000)

AR(2)

−0.35(0.000)

−0.44(0.000)

−0.40(0.000)

R2

0.9996

0.9995

0.9995

R2 Adjusted

0.9995

0.9995

0.9995

Standard regression error

699.23

250.20

743.72

Durbin-Watson

1.82

1.89

1.83

Remarks

374

340

352

as an explanatory part of this phenomenon, it is correct to say that the impact is low (p = 0.1075; t = 1.6) as well as its probability.

4 Discussions and Conclusions NCD have had a relevant increase with economically active population, therefore, diseases are worsening in the world. With mathematical models like used in this research helps to understand the relationship between the economic factors considered in the health sector with chronic diseases. The labor variables like productivity, working productivity as well as economic variables like investment in health and the out-of-pocket budget affect are factors that impact on the NCD and affect the mortality. It’s notable that the impact on mortality derived from diseases by diabetes mellitus and cardiovascular have two important effects, on the one hand, the approach of eating habits and physical activity, which impacts on the disease. It was observed that the feeding of products derived from wheat, corn, as well as sweetening drinks, have a significance in mortality. From the economic point of view, the reduction of investments in the health sector impacts cardiovascular disease. The literature shows that arterial hypertension is

142

S. A. Domínguez-Miranda and R. Rodriguez-Aguilar Table 4 PLS results—neoplasm mortality

Dependent variable: Mortality from diseases caused by neoplasm diseases Independent variables

Model 1

Productivity per hour

−19.93(0.026)

Total population

144.98(0.000)

Incidence of Neoplasm diseases

0.012(0.000)

AR(1)

1.56(0.000)

AR(2)

−0.55(0.000)

R2

0.9996

R2 Adjusted

0.9995

Standard regression error

332.68

Durbin-Watson

2.03

Remarks

360

Table 5 PLS results—neoplasm incidence Dependent variable: Increase in the incidence of diseases caused by Neoplasm diseases Independent variables

Model 1

Alcohol consumption

0.007(0.068)

Working population

0.024(0.002)

Meat consumption

0.0005(0.041)

AR(1)

2.02(0.000)

AR(2)

−1.27(0.000)

AR(3)

0.24(0.000)

R2

0.9999

R2 Adjusted

0.9999

Standard regression error

757.95

Durbin-Watson

1.94

Remarks

238

relevant for cardiovascular disease [16], an element that is corroborated by the analyzes performed, however an element to denote is that the prevalence in hypertension shows a negative significance, it is highly probable that this is because when there is control of hypertension, mortality decreases. In the case of Neoplasm diseases, although there are essential elements to discuss again such as investment in health and food, they do not show high significance, so other variables should be considered such as genetic load, ethnological profile, or exposure

The Economic Dimensions of the Non-communicable Diseases

143

Table 6 PLS results—respiratory diseases Dependent variable: Mortality from diseases caused by respiratory diseases Independent variables

Model 1

Total population

111.32(0.0)

Exposure to pollution

Model 2 129.94(0.002)

Incidence of respiratory diseases

0.005(0.1075)

AR(1)

2.02(0.000)

AR(2)

−1.27(0.000)

R2

0.9999

0.9996

R2 Adjusted

0.9999

0.9996

Standard regression error

757.95

998.70

Durbin-Watson

1.94

2.23

Remarks

238

136

to various elements as sun or radiation. Further research in this area is necessary to statistically understand the behavior of the disease. In mortality derived from respiratory diseases, only one relevant element with significance was found, which is exposure to pollution, it is relevant to find additional indicators that show exposure to other environments or in contact with dangerous elements in the work environment. This indicator gives us the opportunity for further research into elements of exposure to which workers and people may have contact and can generate problems to increase respiratory disease, and, therefore, death. Finally, in all models is found as a variable a significant autoregressive model (p = 0.000), explaining that the effect of mortality is repeated in 2, 3 and up to 8 years, this shows us the need to invest in health because the effect that occurs now by the improvement in prevention or early diagnosis is reflected in 2, 3 or up to 8 years. More research is necessary to elucidate the factors related with NCD in each of the countries and more precisely the economic level to which they belong to be able to better evaluate the disease, however, with the information found, it is possible to recommend appropriate strategies for the early diagnosis and control of NCD to reduce mortality as well as policies to motivate people and employees to improve nutrition, on the other hand investments in the health sector, although they have been a constant recommendation, this time it is observed in the analyses, mainly in cardiovascular disease, which represents the leading cause of death in the world, as a crucial element to minimize the problem of mortality.

References 1. Cordova-Villalobos, J.A., et al.: Chronic noncommunicable diseases in Mexico: epidemiological overview and comprehensive prevention. Salud Pública de México 50(5), 419–427 (2008). https://doi.org/10.1590/s0036-36342008000500015 2. WHO, World Health Organization: The top 10 causes of death (2020). https://www.who.int/ news-room/fact-sheets/detail/the-top-10-causes-of-deat

144

S. A. Domínguez-Miranda and R. Rodriguez-Aguilar

3. OPS, Pan American Health Organization: Leading causes of mortality and disability (2022). Last accessed 2023/29/03. https://www.paho.org/es/enlace/causas-principales-mortalidaddiscapacidad 4. Jaspers, L., et al.: The global impact of non-communicable diseases on households and impoverishment: a systematic review. Eur. J. Epidemiol. 30(3), 163–188 (2014). https://doi.org/10. 1007/s10654-014-9983-3 5. Chaker, L., et al.: The global impact of non-communicable diseases on macro-economic productivity: a systematic review. Eur. J. Epidemiol. 30(5), 357–395 (2015). https://doi.org/ 10.1007/s10654-015-0026-5 6. Rodríguez-Aguilar, R., Rivera-Peña, G., Ramírez-Pérez, H.X.: Household expenditure in health in Mexico, 2016. In: Vasant, P., Zelinka, I., Weber, G.-W. (eds.) ICO 2019. AISC, vol. 1072, pp. 662–670. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-33585-4_64 7. Rashida, M., Iffath, F., Karim, R., Arefin, M.S.: Trends and techniques of biomedical text mining: a review. In: Vasant, P., Zelinka, I., Weber, G.-W. (eds.) ICO 2021. LNNS, vol. 371, pp. 968–980. Springer, Cham (2022). https://doi.org/10.1007/978-3-030-93247-3_92 8. Wang, Y., Wang, J.: Modelling and prediction of global non-communicable diseases. BMC Public Health 20, 1–13 (2020). https://doi.org/10.1186/s12889-020-08890-4 9. Allen, L.N., Wigley, S., Holmer, H.: Implementation of non-communicable disease policies from 2015 to 2020: a geopolitical analysis of 194 countries. Lancet Glob. Health 9(11), e1528–e1538 (2021). https://doi.org/10.1016/S2214-109X(21)00359-4 10. Perazzi, J.R., Merli, G.O.: Panel data regression models and their application in the evaluation of impacts of social programs. Telos 16(1), 157–164 (2014). https://www.redalyc.org/pdf/993/ 99330402007.pdf 11. IHME—Institute for Health Metrics and Evaluation: Global Burden of Disease Study 2019, Results (2020). https://vizhub.healthdata.org/gbd-results/ 12. WHO, World Health Organization: The Global Health Observatory (2022). Last accessed 2023/03/29. https://www.who.int/data/gho/data/indicators/indicator-details/GHO/ mean-bmi-(kg-m-)-(age-standardized-estimate 13. OECD: Working Age Population (Indicator) (2022). https://doi.org/10.1787/d339918b-en. https://data.oecd.org/pop/working-age-population.htm 14. FAO: Food Balances (2022). Last accessed 2023/03/29. https://www.fao.org/faostat/en/#data/ FBSH 15. Wooldridge, J.M.: Introductory econometrics: a modern approach. Cengage Learning (2015) 16. de la Torre Díez, I., Garcia-Zapirain, B., Méndez-Zorrilla, A., López-Coronado, M.: Monitoring and follow-up of chronic heart failure: a literature review of eHealth applications and systems. J. Med. Syst. 40(7), 1–9 (2016). https://doi.org/10.1007/s10916-016-0537-y

Re-strengthening of Real Sized RC Beams Subjected to Corrosion Using Glass Fiber Reinforced Polymer Sheets Sunil Garhwal1,2(B) , Shruti Sharma2 , Sandeep Kumar Sharma3 , Anil Garhwal2 , and Anirban Banik4 1 Department of Civil Engineering, MMEC, Maharishi Markandeshwar University, Mullana,

Ambala 133203, India [email protected] 2 Department of Civil Engineering, Thapar Institute of Engineering and Technology, Patiala, Punjab 147004, India [email protected] 3 Department of Mechanical Engineering, Thapar Institute of Engineering and Technology, Patiala, Punjab 147004, India [email protected] 4 Department of Civil Engineering, National Institute of Technology Sikkim, Ravangla, South Sikkim 737139, India

Abstract. Reinforced concrete (RC) structures have been an integral part of today’s civilization. The strength and stability of a RC structure come to stake when it is exposed to corrosion. Thus the focus of the research is to inspect real sized Reinforced Concrete (RC) beams subjected to various level of corrosion and re-strengthening them using Glass Fiber Reinforced Polymer (GFRP). With increase in the corrosion level the RC beams indicated increase in brittleness and reduced ductility as the steel area is reduced drastically. A substantially increase in ultimate load and deflection characteristics is observed when corroded beams were repaired using GFRP. This Research facilitate the use of GFRP sheets to rehabilitate structures subjected to savior damage due to corrosion. Keywords: Corrosion · GFRP repairing · Deflection · Ultimate load · Ductility

1 Introduction The most often used building material is cement concrete that has been strengthened with steel bars. The vulnerability of these buildings to environmental attack is a serious issue that can significantly lower their strength and lifespan [1]. When it’s humid, air pollution seeps through the concrete cover and corrodes the steel reinforcing. The corrosion products that arise take up space that is several times more than that of the steel [2–8]. Tensile forces imposed on by the increased volume induce cracking, delamination, and spalling in the concrete. As a result, the corrosion is accelerated and the reinforcements are vulnerable to immediate exposure to aggressive agents. In addition to being poor in © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 145–152, 2024. https://doi.org/10.1007/978-3-031-50158-6_15

146

S. Garhwal et al.

aesthetic, it significantly compromises the concrete structure. Moreover, it also effects the interfacial bond between steel and concrete which ultimately influence the ultimate load carrying capacity of structure. Fiber reinforced polymer have emerged as an efficient and effective repairing material which not only strengthens the RC structure but also act as protecting shield to aggressive agents and prevent further corrosion [9–15]. In the present study real size RC beams were cast and were subjected to various level of corrosion and were compared with the rehabilitated beams. We would like to draw your attention to the fact that it is not possible to modify a paper in any way, once it has been published. This applies to both the printed book and the online version of the publication. Every detail, including the order of the names of the authors, should be checked before the paper is sent to the Volume Editors.

2 Experimental Program and Methodology 2.1 Specimen Details Seven RC beams of size 127 mm × 227 mm × 4100 mm were cast for the experimental investigation using a design mix proportion of 1:1.5:3 (cement, sand, aggregate). As indicated in Table 1, the damaged beams advanced to various stages of corrosion before being further repaired by GFRP wrapping. The procedure for accelerated corrosion of beams and the repair method using GFRP have been detailed in the following parts. Table 1. Nomenclature used for specimens Specimen

(Corroded at constant voltage of 10 V)

C-0

Control

C-20

Corroded for 20 days

C-30

Corroded for 30 days

C-40

Corroded for 40 days

H-20

C-20 repaired with GFRP

H-30

C-30 repaired with GFRP

H-40

C-40 repaired with GFRP

2.2 Accelerated Corrosion of RC Beams As the wet and dry processes takes much longer time, the RC beams in this study were exposed to expedited chloride-induced corrosion utilising the impressed current method. The centre 1.5 m of the beam had stainless steel wire mesh wrapped around it, and a water tank system was used to continually provide 3.5% NaCl solution (Fig. 1). The corrosion of RC beams was carried out at a constant voltage of 10 V. The negative end of the power source was connected to the wire mess acting as cathode. The positive terminal of the power supply was linked to the Main steel acting as anode. Each corrosion cycle lasted 20, 30, and 40 cycles on six identical RC beams.

Re-strengthening of Real Sized RC Beams Subjected to Corrosion

147

Fig. 1. Accelerated impressed current corrosion set up

2.3 Repairing of Corroded Beams RC beams subjected to varying level of corrosion damaged (C-20, C-30, and C-40) was Re-strengthened using GFRP sheets. The re-strengthening was conducted in five steps. In first step concrete damaged due to tensile pressure exerted by corrosion product was removed by tapping. In step two the exposed surface was cleaned and two layers of Dr. Fixit Epoxy 211 was applied so as a proper bond between old concrete and new concrete can be achieved. Further in step three micro-concrete placed layer by layer. In step four the surface of harden micro concrete was grinded to make it smooth so to prevent any air gap between concrete surface and GFRP sheets. In the last step two layers of GFRP sheets were applied at the bottom of the beam and a U-wrap in the middle portion to provide confinement pressure on micro-concrete (Fig. 2).

3 Results 3.1 Load Deflection Characteristics The corresponding loads and deflections were measured at mid-span on corroded beams and their GFRP repairs after they had been loaded and tested in two-point bending to failure. It was demonstrated that when corrosion levels increased, the load-bearing

148

S. Garhwal et al.

Fig. 2. Repairing of damaged RC beams with GFRP

capacity declined noticeably, but the GFRP-repaired corroded beams had dramatically enhanced load-bearing capacity that was similar to the control beams. RC beam (C-0) with no damage was subjected to flexural loading and load deflection parameters were recorded and kept as a base line parameter. The beam successfully undertook a load of 49.56 KN and deflection of 121 mm. For beams corroded to varying degrees, a significant decrease in load-bearing capacity and deflection at the mid-span is observed with increasing corrosion. For the C-20 beam, maximum load is significantly reduced to 31.44 KN and mid-span deflection of 52.91. For the severely corroded beam (C-30), a sudden mid-span failure with corrosion cracks integrated into large vertical cracks is observed at a maximum deflection of 47.59 mm and a load of 22 kN. Finally, a severely degraded C-40 beam with numerous vertical corrosion cracks along with longitudinal cracks suddenly failed at a very small load of 17.43 kN with a mid-span deflection of 31.48 mm. Therefore, the reduction in ultimate load and mid-span deflection with increasing corrosion levels is alarming and should be carefully addressed. In addition, girders corroded to varying degrees and repaired with micro-concrete and GFRP coating showed significantly improved load-bearing capacity compared to the original corroded beams. The H-20 beam reported an increase in ultimate load of 46.27 KN and major increase in ultimate deflection almost equal to health specimen that is 101.85 mm. RC beam repaired after 30 and 40 days of corrosion that is H-30 and H-40 reported an increase in load and deflection. H-30 showed an increase in ultimate load as compared to C-30 from 22 KN to 43.15 KN and Deflection from 47.59 to 95.83 mm. Similarly, H-40 showed an impressive improvement in ultimate load of 32 KN and deflection of 77.3 mm. This increase in ultimate load and deflection of repaired specimen clearly demonstrate the effectiveness of GFRP sheets in rehabilitating the structures (Fig. 3) (Table 2; Figs. 4 and 5).

Re-strengthening of Real Sized RC Beams Subjected to Corrosion

149

Fig. 3. Load-deflection characteristics of corroded and repaired beams, a. Load deflection— healthy beam

150

S. Garhwal et al.

Fig. 3. (continued)

Fig. 4. Load characteristics of corroded and repaired beams

4 Conclusions This study focuses on evaluating the performance of real sized RC beams corroded to various levels and repaired with GFRP. Following major conclusions have been drawn from the study: 1. Visible inspections reveal that RC beam corrosion damage is accelerating, growing the huge longitudinal fractures and causing transverse cracking of the rust-covered beams. 2. The results of bending tests on corroded RC beams demonstrate that as corrosion level rises, maximum load capacity and maximum deflection decrease. 3. Increase in ultimate load and deflection is observed when repaired using GFRP sheets.

Re-strengthening of Real Sized RC Beams Subjected to Corrosion

151

Fig. 5. Deflection characteristics of corroded and repaired beams

Table 2. Ultimate load and deflection of corroded & GFRP repaired specimen Specimen

Load (KN)

Deflection (mm)

Healthy

49.56

121.053

C-20

31.44

52.91

C-30

22.015

47.59

C-40

17.43

31.486

H-20

46.275

101.857

H-30

43.15

95.833

H-40

32.025

77.316

References 1. Elsener, B., Angst, U.: Corrosion inhibitors for reinforced concrete. In: Science and Technology of Concrete Admixtures (2016) 2. Garhwal, S., Sharma, S., Sharma, S.K.: Acoustic emission monitoring of RC beams corroded to different levels under flexural loading. Arab. J. Sci. Eng. 46(5), 4319–4335 (2020). https:// doi.org/10.1007/s13369-020-04930-8 3. Garhwal, S., Sharma, S., Sharma, S.K.: Monitoring the flexural performance of GFRP repaired corroded reinforced concrete beams using passive acoustic emission technique. Struct. Concr. (2020). https://doi.org/10.1002/suco.202000247 4. Garhwal, A., Sharma, S., Danie Roy, A.B.: Performance of expanded polystyrene (EPS) sandwiched concrete panels subjected to accelerated corrosion. In: Structures, vol. 43, pp. 1057–1072 (2022)

152

S. Garhwal et al.

5. Sharma, S., Mukherjee, A.: Monitoring corrosion in oxide and chloride environments using ultrasonic guided waves. J. Mater. Civ. Eng. (2011). https://doi.org/10.1061/(ASCE)MT.19435533.0000144 6. Sharma, A., Sharma, S., Sharma, S., Mukherjee, A.: Investigation of deterioration in corroding reinforced concrete beams using active and passive techniques. Constr. Build. Mater. (2018). https://doi.org/10.1016/j.conbuildmat.2017.11.165 7. Shan, H., Xu, J., Wang, Z., Jiang, L., Xu, N.: Electrochemical chloride removal in reinforced concrete structures: improvement of effectiveness by simultaneous migration of silicate ion. Constr. Build. Mater. (2016). https://doi.org/10.1016/j.conbuildmat.2016.09.137 8. Ohtsu, M., Tomoda, Y.: Phenomenological model of corrosion process in reinforced concrete identified by acoustic emission. ACI Mater. J. (2008). https://doi.org/10.14359/19764 9. Gadve, S., Mukherjee, A., Malhotra, S.N.: Corrosion protection of fiber-reinforced polymerwrapped reinforced concrete. ACI Mater. J. (2010). https://doi.org/10.14359/51663860 10. Siddika, A., Al Mamun, M.A., Alyousef, R., Amran, Y.H.M.: Strengthening of reinforced concrete beams by using fiber-reinforced polymer composites: a review. J. Build. Eng. (2019). https://doi.org/10.1016/j.jobe.2019.100798 11. Mukherjee, A., Boothby, T.E., Bakis, C.E., Joshi, M.V., Maitra, S.R.: Mechanical behavior of fiber-reinforced polymer-wrapped concrete columns—complicating effects. J. Compos. Constr. (2004). https://doi.org/10.1061/(ASCE)1090-0268(2004)8:2(97) 12. Mukherjee, A., Gadve, S., Malhotra, S.: Active protection of FRP wrapped reinforced concrete structures against corrosion. In: Concrete Solutions (2009) 13. Kreit, A., Al-Mahmoud, F., Castel, A., François, R.: Repairing corroded RC beam with nearsurface mounted CFRP rods. Mater. Struct. Constr. (2011). https://doi.org/10.1617/s11527010-9693-6 14. Triantafillou, T.C., Antonopoulos, C.P.: Design of concrete flexural members strengthened in shear with FRP. J. Compos. Constr. (2000). https://doi.org/10.1061/(ASCE)1090-0268(200 0)4:4(198) 15. Soudki, K.A., Sherwood, T., Masoud, S.: FRP repair of corrosion-damaged reinforced concrete beams. In: Proceedings of the 3rd Fiber International Congress—2010 (2002)

Optimization of the Lubricating and Cooling Fluid Composition I. Yu. Ignatkin1 , P. Kazantsev Sergey1 , D. M. Skorokhodov1(B) , N. V. Serov1 , T. Kildeev2 , A. V. Serov1 , and A. Anisimov Alexander1 1 Russian State Agrarian University—Moscow Timiryazev Agricultural Academy,

Moscow 127550, Russia {ignatkin,d.skorokhodov}@rgau-msha.ru 2 Bauman Moscow State Technical University, Moscow 105005, Russia

Abstract. The article deals with the issue of the durability of a metal-cutting tool increasing. The main advantages of oil lubricants and coolants are compared. The hypothesis of the taps durability increase due to the usage of an oil-based coolant with the use of a metal-coating additive is being investigated in the study. The metal-coating additive “Valena” was used to increase the taps durability in the conditions of oil-based lubricating and cooling liquid. Industrial oil I-30 was used as a lubricating and cooling medium for control samples. Studies of the durability period were carried out on machine taps M10x1 made of steel grade P6M5. Thread cutting was performed by the vertical drilling machine 2H118. During the experiment, when the workpiece made of Steel 40X material was processed, the appearance of a silver film on the surface of the Valena solution was observed. When the steel St3 was processed, the silver film was not observed. Keywords: Durability · Wear · Oil lubricants · Metal-cutting tool · Tap · Additive · Coolant · Friction

1 Introduction The production and renovation of equipment is always accompanied by mechanical surface treatment [1], which leads to the wear of the tool. It indicates the urgency of the durability of metal-bearing tools increasing problem. The solution of this problem can be carried out by technological and constructive ways, for example, through the use of the new tool materials and wear-resistant coatings applied to tools by surfacing or welding [2, 3], gas-dynamic or gas-thermal methods [4– 6], PVD and CVD methods, as well as replacing metal workpieces (objects of mechanical processing) with more technological and durable products made of polymer composite materials [7, 8]. Considering all the listed methods, it is advisable to describe the reasons of the cutting tool wear. Concerning the process of wear, it should be noted that there are permissible (relatively smooth change in size and shape) and unacceptable (damage) types of wear [9]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 153–162, 2024. https://doi.org/10.1007/978-3-031-50158-6_16

154

I. Yu. Ignatkin et al.

In terms of the cutting tool durability increasing, the permissible types of wear are of the greatest interest. According to the physico-chemical mechanism the impact on the tool can be divided into: abrasive (mechanical), oxidative (chemical-mechanical), wear of films of non-oxygen origin (mechanochemical) [10]. Oxidative wear (OW). Occurs under conditions of friction in an air or other oxygencontaining medium. The oxides form a protective layer on the friction surface, protecting it from setting, so that the OW speed is relatively small. However, under the influence of plastic deformation, oxides are destroyed and, having a high hardness, can act as an abrasive. The destroyed oxides are quickly restored, which is acceleratev by epy elevated temperatures in the friction zone [11]. Another form of OW is formed by the oxide films on the surface of rubbing parts with their subsequent removal. The formed structures have significantly greater hardness and brittleness than the base metal. The formed oxide films have a clear partition boundary with the base material, and the bond strength is low. The combination of high brittleness and low strength of the joint leads to intensive destruction and the formation of new films, while the surface acquires a heterogeneous structure. The wear of the films of non-oxygen origin involves the formation of sulfide, phosphate, nitride, servovite and other films on the surface with their subsequent destruction. Hydrogen wear (HW) consists in the diffusion of atomic hydrogen into the material of friction pairs and the subsequent embrittlement of interfacial boundaries, grain boundaries, the development of microcracks. Sources of atomic hydrogen surround the friction zone, mechanical work, elevated temperature contribute to the active release of atomic hydrogen from organic compounds of fuels and lubricants, plastics, and water. At the same time, two fundamentally different types of hydrogen wear are distinguished: by dispersion and by destruction. In the first case, as a result of plastic deformation, the hydrogen-embrittled material is fragmented, destroyed with the formation of a fine metal powder. At the same time, the surface doesn’t show the visible signs of wear. In case of destruction, hydrogen penetrates into the pores of the metal and combines into molecules, under the pressure of the friction pair, the pores close, the pressure inside the pores increases and destruction occurs along all microcracks, a metal flake is instantly separated to form a cavity [12]. Abrasive wear (AW) occurs as a result of interaction with abrasive particles having a significant superiority in hardness. In this case, abrasive particles are usually divided into fixed, tangentially in contact with the surface; loose abrasive; free particles located in the gap of the friction pair; free particles carried by the flow of liquid or gas [13]. In order to avoid premature failure for the reasons described above, it is necessary to strengthen the tool, modify the design or protect (in whole or in part) the tool from the effects of negative factors. All available hardening methods are divided into 6 main classes: 1. 2. 3. 4.

hardening with the formation of a film on the surface; with a change in the chemical composition of the surface layer; with a change in the structure of the surface layer; with a change in the energy reserve of the surface layer;

Optimization of the Lubricating and Cooling Fluid Composition

155

5. with the change of the microgeometry of the surface and the riveting; 6. with a change in the structure over the entire volume of the material. The improvement of the structure by heat treatment (cold treatment, steam treatment) is also applied. Optimization of geometric parameters is carried out in the direction of macrogeometry (angles, chip breakers, rags, reinforcing chamfers, etc.) and microgeometry (regular microrelief, surface roughness, etc.). Improving the surface quality of the tool can be carried out, for example, by fine-tuning. For increase of the durability of the cutting edge it should be adjusted. Sharpening and finishing of cutting edges, especially with diamond circles, makes it possible to increase the average durability of a number of tools, especially finishing ones, by 2–3 times or more, since it improves the surface quality of the tool, and, consequently, the working conditions of the cutting part of the tool. Special attention should be paid to the use of lubricating and cooling process media (LCPM) [14]. LCPM perform cooling, lubricating and washing functions, which makes it possible to carry out metalworking with greater productivity, achieve higher surface quality, and wash out chips from the cutting zone [15]. LCPM are presented in the following forms: – – – –

gaseous LCPM; plastic LCPM (technological lubricants); liquid LCPM (LCF, or lubricating and cooling fluids); solid LCPM.

Gaseous LCPM are represented by inert or neutral gases (helium, argon, nitrogen), active gases (oxygen, carbon dioxide, air), there are also examples of additional activation of gases, for example, by ionization. Activation allows intensifying of the protective films formation on the surfaces of interacting parts. However, the complexity of application, low heat capacity, lack of detergent properties limit the use of gaseous LCPM in practice [16, 17]. Plastic LCPM are applied during manual low-productivity processing or in highly loaded periodic types of processing. The main difficulties are caused by the complexity of the supply (withdrawal) of the material to the processing zone, collection, cleaning, cyclic use. Cooling and washing effects from the use of plastic STS are practically absent. Mineral materials of layered structure (graphite, molybdenum disulfide, talc, mica), soft metals (lead, copper, tin) or organic compounds (wax, soap, solid fats, polymers) are used as solid LCPM. These lubricating and cooling compounds are applied as coatings to the treated surface or tool. They are used at high loads and temperatures when the use of other types of LCPM is impossible or difficult. Under normal processing conditions, solid LCPM is usually not used due to the low efficiency of the heat removal and the complexity of the application [10, 13]. The most common LCPM currently are lubricating and cooling fluids (LCF), which are able to cope with the tasks of cooling, lubrication and washing out of cutting products from the processing zone. According to the composition, they are divided into the following types:

156

I. Yu. Ignatkin et al.

• rapidly evaporating; • oil; • water-based. Oil lubricants and coolants are made on the basis of mineral oils. The composition of these LCF may also include antifriction, antiburring, anti-wear, anti-foam and antifog additives; corrosion inhibitors, antioxidants. The physico-chemical properties of oil LCF and their characteristics affecting the process of friction and wear of surfaces are determined, first of all, by the base oils included in the material composition. As a base for oil LCF, well purified mineral paraffin or naphthenic oils, low-viscosity extracts of selective purification, as well as mixtures of several mineral oils can be used. The oil content in the coolant of this class is usually 60.95% by the weight. Synthetic oils have a fairly high cost. They can be used in oil lubricants as additives. Oils without additives are used as coolant in light modes of soft metals cutting (copper, brass, bronze, magnesium, carbon steels). In severe cutting conditions of hardto-process steels and alloys, oil lubricants without the addition of an additive package are usually ineffective [12]. The concentration of antifriction additives in oil coolants ranges from 5 to 25%. They are usually based on the organic or polymeric unsaturated fatty acids, their esters, vegetable oils and fats. Extreme pressure additives in oil coolants are substances containing sulfur, chlorine, phosphorus. The most common among them are sulfides, poly–sulfides, blackened fats, chlorinated paraffin. Their content in the composition of the material varies from 0.5 to 20% and depends on the purpose and conditions of application of oil coolants. The content of anti-wear additives (polymer fatty acids, dialkyl phosphates or blackened fats) in coolant usually ranges from 0.5 to 5%. It depends on the purpose of the liquid. Polyolefins or atactic polypropylene are used as anti-fog additives in coolant. These substances are introduced into lubricants in an amount of 0.5–3% to reduce the formation of oil mist. Dimethylselicone polymers have become the most widespread among anti-foam additives. Their concentration in oil coolants is 0.0005–0.001% [10, 11]. Under the influence of the mineral oils oxidation products, additives, their decomposition products, corrosion can forme on the parts of the equipment and the processed nodes. Corrosion inhibitors, which are introduced into the composition of oil coolant, prevent its occurrence. The tendency to corrosion in different structural materials can vary significantly, so inhibitors are selected depending on the application of coolant. In some cases, additives for improving lubricating properties can be quite effective corrosion inhibitors: unsaturated fatty acids, disulfides, aminophosphates. Oil LCF have a whole range of advantages compared to other types of LCPM: • • • • • •

they provide longer operation of the cutting tool; have excellent lubricating properties; better protect the processed metal and tools from corrosion; chips and tool wear products are washed out of the cutting zone more efficiently; can be used in centralized lubrication systems; subject to recycling, cleaning and reuse.

However, oil coolants also have a number of disadvantages that significantly narrow the scope of their application. They are flammable, have high evaporation, relatively

Optimization of the Lubricating and Cooling Fluid Composition

157

low cooling properties and low thermal stability. In addition, the use of oil coolants is a rather expensive way to lubricate and cool the cutting tool [17–19]. Sulfofresol has proven itself well—a mineral oil of medium viscosity, which includes sulfur. When processing under the influence of high temperatures, protective sulfide films are formed on the surfaces of the workpiece and the tool, which significantly reduce the coefficient of friction. At high processing speeds, sulfofresol actively evaporates and smokes. Sulfur is consumed during processing, which reduces the efficiency of LCF, and resinous formations are deposited on the machines. Sulfofresols are toxic. A promising direction in processing is the use of LCF with anti-wear metal-coating additives. The metal-coating additive “Valena” implements the effect of wear-free friction discovered by D.N. Garkunov and I.V.Kragelsky. The essence of the effect is the formation of a protective copper film with a spongy structure on the surface of the rubbing parts. The film thickness is about one micrometer. In this paper, it is proposed to conduct the study of the hypothesis of increasing the durability of taps due to the use of an oil-based coolant with the supplementation of a metal-coating additive.

2 Materials and Methods Studies of the durability period were carried out on machine taps M10x1 made of steel grade P6M5. To test the hypothesis of increasing the resistance of taps under the conditions of using an oil-based lubricating and cooling liquid with the use of a metal-coating additive “Valena”, an experimental solution was prepared: industrial oil I-30 with an additive “Valena”, the concentration of the solution is 20%. To test the hypothesis, the concentration (20%) is taken as the arithmetic mean of the manufacturer’s recommendations (10–30%). If the hypothesis is confirmed, it will be advisable to conduct research on optimizing the composition of LCF. Industrial oil I-30 was used as a lubricating and cooling process medium for control samples. The thread was cut in through holes with a height of 30 mm. The processing material is 40X steel with low-temperature tempering (200 °C). The tested taps cutted the holes until their blunting. According to the results of installation tests, the reference wear with the use of coolant with the additive “Valena” was established. Counting was stopped when comparable wear with the use of pure I-30 was achieved. The degree of wear was monitored through every 5 holes on the Supereyes b008 microscope (Fig. 1) with the ability of the connection with the computer via the usb interface. The ratio of the number of holes produced, provided that the processing modes are the same, is equal to the ratio of the durability periods. k=

z2 T2 = , T1 z1

(1)

where T 1 , T 2 —the period of the tap durability when using the experimental and control LCPM respectively, min.; z1 , z2 —the number of threaded holes produced when using the experimental and control LCPM respectively, pieces.

158

I. Yu. Ignatkin et al.

Fig. 1. Supereyeb008 microscope.

Threading was performed with the vertical drilling machine 2H118 (Fig. 2). The spindle speed n = 350 rpm, which corresponds to a cutting speed of 11 m/min. The feed corresponds to the pitch of the thread being cut (1 mm).

Fig. 2. Column drilling machine 2H118.

3 Results According to the results of the installation tests, the condition of the taps after processing 100 holes with the use of LCF t with the additive “Valena” was taken as the reference wear (Fig. 3). The test results are shown in Table 1 (Fig. 4).

4 Discussion In the process of drilling holes, the appearance of a silver film on the surface of the Valena solution was observed during the processing a workpiece made of Steel 40X material (Fig. 5). When working with steel St3, this film was not observed. It can be assumed that the alloying element chromium is isolated from the material. The discovered effect requires further research.

Optimization of the Lubricating and Cooling Fluid Composition

159

Table 1. Test results. № 1 2

Cutting fluid 20% «Valena»

Industrial oil I-30

100

30 30

3

35

Average value

32

Fig. 3. Tap cutting edge: a original, b worn

Fig. 4. Resistance taps with different cutting fluid: Z 1 —the number of machined holes with a tap with LCF 1; Z 2 —the number of machined holes tap with LCF 2

160

I. Yu. Ignatkin et al.

In further work, we plan to clarify the optimal concentration of the additive in the composition, as well as the series of tribotechnical tests of solutions on a friction machine. As well as a study on increasing the cooling and washing abilities of the resulting coolant.

Fig. 5. Silver film formation.

5 Conclusion When threading in 40X Steel, the use of a 20% solution of the Valena metal-coating additive in industrial oil I-30 in the quality of LCPM allows the increase of the period of taps resistance by more than 3 times. To obtain the maximum durability period, it is necessary to investigate the effect of the concentration of the multifunctional additive “Valena” and the cutting speed on the specified parameter, which is planned to be done in the future. The research was conducted with the financial support of the K.A. Timiryazev Russian State Agrarian University-Moscow Agricultural Academy program “Scientific internship” (the theme of the project “Research of structural materials and nanostructured functional coatings for agricultural machinery and equipment”) as part of the implementation of the University development program “Agroproyv-2030” of the academic strategic leadership program “Priority-2030”.

References 1. Kononenko, A.S., Khabbatullin, R.R.: Theoretical substantiation of the conditions for the applicability of deformationless fixation by means of a polymer glue for workpieces during their mechanical processing on a milling machine with computer numerical control. In: Polymer Science. Series D, vol. 15, no. 4, pp. 523–528 (2022). https://doi.org/10.1134/S19954 21222040141 2. Latypov, R.A., Serov, A.V., Ignatkin, I.Y., Serov, N.V.: Utilization of the wastes of mechanical engineering and metallurgy in the process of hardening and restoration of machine parts. Part 1. In: Metallurgist, vol. 65, no. 5–6, pp. 578–585 (2021). https://doi.org/10.1007/s11015-02101193-y 3. Latypov, R.A., Serov, A.V., Ignatkin, I.Y., Serov, N.V.: Utilization of the wastes of mechanical engineering and metallurgy in the process of hardening and restoration of machine parts. Part 2. In: Metallurgist, vol. 65, no b/n, pp. 689–695 (2021). https://doi.org/10.1007/s11015-02101206-w

Optimization of the Lubricating and Cooling Fluid Composition

161

4. Zayatzev, A.N., Lukianova, A.N., Demoretsky, D.A.: Assessment of shear bond strength of thermal spray coatings by applying prismatic samples. Solid State Phenom. 337, 35–41 (2022). https://doi.org/10.4028/p-297el3 5. Zayatzev, A.N., Alexandrova, Y.P.: Effect of elevated temperatures on tribological characteristics of a plasma-sprayed Al2 O3 coating under sliding conditions without lubrication. J. Friction Wear 41(3), 242–246 (2020). https://doi.org/10.3103/S1068366620030174 6. Zayatzev, A.N., Alexandrova, Y.P.: Reduction of shear stresses in friction units with an electrical insulation coating of the ITER blanket modules. J. Mach. Manuf. Reliab. 49(9), 763–769 (2020). https://doi.org/10.3103/S1052618820090137 7. Kononenko, A.S., Solovyeva, A.A., Komogortsev, V.F.: Theoretical determination of the minimum thickness of a polymer layer providing ensured protection of a shaft–bearing joint from fretting corrosion. In: Polymer Science. Series D, vol. 13, № 1, pp. 45–49 (2020). https:// doi.org/10.1134/S1995421220010116 8. Kononenko, A.S., Ignatkin, I.Y., Drozdov, A.V.: Recovering a reducing-gear shaft neck by reinforced-bush adhesion. In: Polymer Science, Series D, vol. 15, no 2, pp. 137–142 (2022). https://doi.org/10.1134/S1995421222020113 9. Fanidi, O., Kostryukov, A.A., Shchedrin, A.V., Ignatkin, I.Y.: Comparison of analytical and experimental force in cylindrical workpiece drawing process. In: Tribology in Industry, vol. 43, no 1, pp. 1–11 (2021). https://doi.org/10.24874/ti.1000.11.20.02 10. Garkunov, D.N.: Tribotehnika (konstruirovanie, izgotovlenieijekspluatacijamashin) [Tribotechnology (de-sign, manufacture and operation of machines)], Publ., Moscow «MSHA» 632 p (2002) 11. Sorokin, G.M.: Tribologijastalejisplavov [Tribology of steels and alloys], p. 314. Publ. Nedra, Moscow (2000) 12. Suranov, G.I.: O mehanizmesnizhenijavodorodnogoiznashivanijadetalejmagnitnojobrabotkoj [On the mechanism of reducing hydrogen wear of parts by magnetic treatment], Jeffektbezyznosnostiitribotehnologii. [The effect of fatigue and tribotechnology.], no. 2, pp. 27–31 (1992) 13. Garkunov, D.N., Kornik, P.I.: Vidytrenijaiiznosa. Jekspluatacionnyepovrezhdenijadetalejmashin [Types of friction and wear. Operational damage to machine parts]. Publ., Moscow «MSHA», 344 p (2003) 14. Skorokhodov, D., Krasnyashchikh, K., Kazantsev, S., Anisimov, A.: Theory and methods of means and modes selection of agricultural equipment spare part quality control. In: Engineering for Rural Development: 19, Jelgava, 20–22 ma 2020 goda. Jelgava, pp. 1140–1146 (2020). https://doi.org/10.22616/erdev.2020.19.tf274. EDN BKFWAI 15. Erokhin, A., Kazantsev, S., Pastukhov, A., Golubev, I.: Theoretical basis of justification of electromechanical hardening modes of machine parts. In: Engineering for Rural Development: 19, Jelgava, 20–22 ma 2020 goda. Jelgava, pp. 147–152 (2020). https://doi.org/10.22616/ ERDev.2020.19.TF032. EDN IDCZQE 16. Tikhomirov, D., Kuzmichev, A., Rastimeshin, S., Trunov, S., Dudin, S.: Energy-efficient pasteurizer of liquid products using IR and UV radiation. In: Vasant, P., Zelinka, I., Weber, G.-W. (eds.) ICO 2018. AISC, vol. 866, pp. 178–186. Springer, Cham (2019). https://doi.org/ 10.1007/978-3-030-00979-3_18 17. Pol’cer, G., Firkovskij, A., Lange, I.I.: i dr. Finishnajaantifrikcionnajabezabrazivnajaobrabotka (FABO) iizbiratel’nyjperenos [Finishing anti-friction non-abrasive treatment (FABO) and selective transfer]. Sb. Dolgovech-nost’ trushhihsjadetalejmashin [Collection. The durability of the rubbing parts of machines.], no 4. Publ. Mashinostroenie, Moscow, 316 p (1990) 18. Dorokhov, A., Kirsanov, V., Pavkin, D., Yurochka, S., Vladimirov, F.: Recognition of cow teats using the 3D-ToF camera when milking in the “Herringbone” milking parlor. In: Vasant,

162

I. Yu. Ignatkin et al.

P., Zelinka, I., Weber, G.-W. (eds.) ICO 2019. AISC, vol. 1072, pp. 128–137. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-33585-4_13 19. Lazar, V.V., et al.: J. Phys.: Conf. Ser. 1679, 042058 (2020). https://doi.org/10.1088/17426596/1679/4/042058

Research of the Dosing Process with the Installation of Magnetic Stimulation of Seeds V. Syrkin1

, S. Mashkov1

, P. Ishkin1

, S. Vasilev1

, and Yu. Daus2(B)

1 Samara State Agrarian University, Uchebnaya St., 2, 446442 Ust-Kinelskiy, Russia 2 Kuban State Agrarian University, Kalinina St. 13, 350044 Krasnodar, Russia

[email protected]

Abstract. Magnetic stimulation of seeds is one of the electrophysical methods that increases the germination and intensity of plant growth. To identify the factors affecting the magnetic field on seeds, an experimental research method was developed using a magnetic seed stimulation unit with a vibrating dispenser. Studies have been conducted on the influence of the frequency of the magnetic field on the process of seed dosing. The minimum supply of Spring Wheat and Amaranth seeds, respectively, was 3.1 kg/h and 1.7 kg/h at a frequency of 10 Hz. The maximum feed was observed at a frequency of 110 Hz—70.1 kg/h and 22.7 kg/h, respectively.The stimulation time at a frequency from 10 Hz to 110 Hz for spring wheat varies in the range from 90 s to 4 s, Amaranth—from 89 s to 7 s. Given the required seed stimulation time, the necessary feed can be set by changing the frequency. Keywords: Magnetic stimulation · Dosing process · Electrophysical methods

1 Introduction Crop production is one of the main directions of the agro-industrial complex, providing the population with food, animal husbandry with feed and raw materials of a number of industrial areas. The intensification of crop production leads to a reduction in cost and an increase in production [1, 2]. The main directions of crop production development are the use of new advanced technologies and equipment, increasing soil fertility, the use of high-quality seed material, etc. [3–5]. The use of high-quality seed provides high germination, the intensity of plant growth, immunity to external factors such as weather conditions and diseases. Over time, the quality of the seed decreases, which leads to a decrease in production. As a result, there is a need to purchase new expensive seeds, while most of the quality indicators of the old seeds remain at an acceptable level [6–8].

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 163–169, 2024. https://doi.org/10.1007/978-3-031-50158-6_17

164

V. Syrkin et al.

A decrease in germination leads to overspending of seed material, which affects the increase in production costs. To increase the germination of seeds, various methods of exposure are used [10–12]. One of the promising methods is the treatment of seeds in a magnetic field. This method is environmentally friendly and does not require large energy costs. There are various devices for magnetic stimulation of plants. However, not all of them provide the main task—the effect on the plant with a uniform magnetic field. Also, some plants operate in a cyclic type, which automates the process and increases the processing time, while flow-type plants can be used in a production line for preparing seeds for sowing [3–8].

2 Materials and Methods When treating seeds with a magnetic field, the main factors affecting seed germination are the parameters of the magnetic field, the stimulation time and the holding time before sowing [9, 10]. It was also found that the seeds are more responsive to a U-shaped magnetic field of different frequencies. At the Department of “Electrification and Automation of the Agroindustrial Complex” of the Samara State Agrarian University, an experimental laboratory installation of magnetic stimulation of seeds of a flow type with a vibrating dispenser was developed (Fig. 1). The installation consists of a frame of a vibrating dispenser 2, a magnetic seed stimulation unit 3, a hopper 4, a power supply and control unit 5. A multimeter 6 was used to set the required frequency. When the unit was operating, a receiving box was placed under the vibrating dispenser [5, 6]. The main feature of the installation design is the magnetic stimulation unit 3, which is a two-circuit branched circuit with inductors on the external branches of the magnetic circuit. In the Central branch of the magnetic circuit 1 (Fig. 2, a) an air gap, which passes through a vertical pipe 2 connecting the hopper 4 (Fig. 1) with vibration pump 2. Vibratory dispenser includes a magnetic circuit 3 (Fig. 1), an induction coil 4, building 5, the pressure limiter seeds 6 and vibrating plates 8. In the process of work, due to the creation of oscillatory movements of the vibrating plates 8 (Fig. 2), the seeds begin to move, move to the edge of the plates and pour off it. At the same time, seeds come from the hopper through the nozzle 2 in their place. In the process of moving along the branch pipe 2, the seeds fall into the zone of the magnetic stimulation unit, where they are exposed to a magnetic field [6]. By changing the frequency of the magnetic field of the dispenser, the vibration frequency of the plates 8 (Fig. 2) changes, as a result of which the seed supply Q changes. The seed supply when moving along the branch pipe will be determined as follows: Q = γ · a · b · v, kg/s, where γ—the seed density, kg/m3 ; a and b—the dimensions of the side walls of the pipe, m; v—the flow rate of seeds in the branch pipe, m/s.

(1)

Research of the Dosing Process with the Installation of Magnetic

165

Fig. 1. General view of the laboratory installation of magnetic stimulation of seeds

Fig. 2. The main components of the installation: (a) plant stimulation zone; (b) vibration dispenser: 1—magnetic core of the seed magnetic stimulation unit; 2—branch pipe; 3—magnetic core of the dispenser; 4—induction coil; 5—dispenser housing; 6—limiter; 7—flap; vibration plate

The flow rate will be equal to: v=

h , m/s, t

where h—the height of the stimulation zone (the height of the electromagnet), m; t—the time of passage of the stimulation zone by seeds, s.

(2)

166

V. Syrkin et al.

Substituting formula 2 into formula 1 we get Q=

γ·a·b·h , kg/s. t

(3)

The supply of the vibrating dispenser is determined as follows: (4) where B—width of plates, m; H—height of the seed layer, m; vd —the average speed of movement of seeds on the vibrating plate, m/s. n—number of plates, pcs. The value of the average speed of movement of seeds on vibrating plates is determined as follows (5) gde k—coefficient depending on the physical and mechanical properties of seeds, determined experimentally; A—vibration amplitude, m; f —frequency of plate vibrations, Hz. Then (6) The movement of seeds along the vibrating plates is carried out due to the lateral pressure of the seeds coming from the branch pipe. As , then we equate the right parts of formulas (3) and (4) we get γ ·a·b·h = B · H · k · A · f · γ · n, t

(7)

After performing the transformations and expressing the time, we get t=

a·b·h . B·H ·k ·A·f ·n

(8)

As a result, the seed stimulation time is inversely proportional to the vibration frequency of the plates. Experimental studies were carried out to determine the effect of the vibration frequency of the vibrating dispenser plates on the seed supply. The variable factor was the frequency of the magnetic field, which directly affects the vibration frequency of the plates. The frequency gradation of the magnetic field was 10, 30, 50, 70, 90, 110, 130 and 150 Hz. For each frequency value, the experiment was carried out with a three-fold repetition. The seed stimulation time range was assumed to be 1 min. The experiments were carried out on wheat and amaranth seeds.

Research of the Dosing Process with the Installation of Magnetic

167

Table 1. Results of studies of the influence of the magnetic f frequency on the supply of the vibrating dispenser Frequency f, Hz Seed supply Q, kg/h. for Spring wheat Seed supply Q, kg/h. for Amaranth 10

3.1

1.7

30

16.3

4.1

50

31.1

7.7

70

47.8

12.7

90

61.7

18.3

110

70.1

22.7

130

61.5

20.1

150

33.2

13.6

3 Results and Discussion The results of the influence of the frequency of vibration of the dispenser plates on its performance are presented in Table 1. When analyzing the results of the experiment, the following data were revealed. That with an increase in frequency in the range from 10 to 110 Hz, the supply of processed seeds increases. The low vibration frequency of the plates had a low incentive for seeds to pour out. At the same time, a linear dependence is observed in this range. In the range from 130 to 150 Hz, a decrease in performance was recorded. This is due to a decrease in the intensity of the vibration action, i.e. the plates made of electrical steel did not have time to return to its original position (Fig. 3). The minimum supply of spring wheat and amaranth seeds, respectively, was 3.1 kg/h and 1.7 kg/h at a frequency of 10 Hz. The maximum feed was observed at a frequency of 110 Hz—70.1 kg/h and 22.7 kg/h, respectively. At the same time, the minimum unevenness of dosing in the range from 10 Hz to 110 Hz of wheat seeds was observed at a magnetic field frequency of 70 Hz and amounted to 1.8%, and for amaranth seeds at a frequency of 90 Hz—1.6%, respectively. The maximum value of uneven dosing in wheat and amaranth seeds was observed at a frequency of 10 Hz and amounted to 4.1% and 3.6%, respectively, which is lower than acceptable values. The stimulation time at a frequency from 10 Hz to 110 Hz for spring wheat varies in the range from 90 s to 4 s, amaranth - from 89 s to 7 s. An additional increase in the stimulation time is achieved by changing the position of the adjustable flap 7 (Fig. 2, b). Considering that the optimal stimulation time for these cultures is in the range from 30 s to 60 s, it can be concluded that the specified range of adjustable time with a large margin corresponds to the stimulation time.

V. Syrkin et al.

Performance Q, kg/h

168

80 60 40 20 0 0

50

100 150 Magnetic field frequency f,Hz

200

0

50

100 150 Magnetic field frequency f,Hz

200

Performance Q, kg/h

a)

25 20 15 10 5 0

b)

Fig. 3. Results of research on the effect of the magnetic field frequency on the supply of a vibrating dispenser: (a) spring wheat seeds; (b) amaranth seeds

4 Conclusion The use of a magnetic seed stimulation unit with a flow-type vibrating dispenser will increase the productivity of the processing process. By changing the frequency of the magnetic field and, as a consequence, the frequency of vibration of the plates of the vibrating dispenser in the range from 10 Hz to 110 Hz, the feed will change, which as a result will affect the stimulation time. For wheat and amaranth seeds, the set feed range largely overlaps the required stimulation time range. At the same time, the uneven dosing of seeds of both crops corresponds to acceptable values. As a result, knowing the necessary time to stimulate a particular culture, it is possible to set the necessary supply using the frequency of the magnetic field.

References 1. Baev, V.I., Yudaev, I.V., Petrukhin, V.A., Prokofyev, P.V., Armyanov, N.K.: Electrotechnology as one of the most advanced branches in the agricultural production development. In: Handbook of Research on Renewable Energy and Electric Resources for Sustainable Rural Development. IGI Global, Hershey, PA, USA (2018) 2. Yudaev, I.V., Daus, Y.V., Kokurin, R.G.: Substantiation of criteria and methods for estimating efficiency of the electric impulse process of plant material. IOP Conf. Ser.: Earth Environ. Sci. 488(1), 012055 (2020)

Research of the Dosing Process with the Installation of Magnetic

169

3. Yudaev, I.V., Ivushkin, D., Belitskaya, M., Gribust, I.: Pre-sowing treatment of ROBINIA PSEUDOACACIA L. seeds with electric field of high voltage. IOP Conf. Ser.: Earth Environ. Sci. 403(1), 012078 (2019) 4. Syrkin, V.A., Vasiliev, S.I., Shishkin, P.A., Smolev, K.S.: Patent No. 204352. RU. Installation for Pre-sowing Seed Stimulation. 2021105476 (2021) (in Russian) 5. Syrkin, V.A., Gridneva, T.S., Ishkin P.A., Fatkhutdinov, M.R.: Device for seed stimulation by pulsed magnetic field. Selskiymekhanizator 6, 28–29 (2019) (in Russian) 6. Vasilev, S.I., Mashkov, S.V., Syrkin, V.A., Gridneva, T.S., Yudaev, I.V.: Results of studies of plant stimulation in a magnetic field. Res. J. Pharm. Biol. Chem. Sci. 9(1), 706–710 (2018) 7. Yudaev, I.V., Daus, Y.V., Gamaga, V.V., Grachev, S.E., Kuligin, V.S.: Plant tissue sensitivity to electrical impulse. Res. J. Pharm. Biol. Chem. Sci. 9(4), 734–739 (2018) 8. Mashkov, S.V., Vasilev, S.I., Fatkhutdinov, M.R., Gridneva, T.S.: Using an electric field to stimulate the vegetable crops growth. Int. Trans. J. Eng. Manage. Appl. Sci. Technol. 11(16), 11A16V (2020) 9. Tokarev, K., et al.: Monitoring and intelligent management of agrophytocenosis productivity based on deep neural network algorithms. Lect. Notes Netw. Syst. 569, 686–694 (2023) 10. Yudaev, I., Eviev, V., Sumyanova, E., Romanyuk, N., Daus, Y., Panchenko, V.: Methodology and modeling of the application of electrophysical methods for locust pest control. Lect. Notes Netw. Syst. 569, 781–788 (2023) 11. Ivushkin, D., et al.: Modeling the influence of quasi-monochrome phytoirradiators on the development of woody plants in order to optimize the parameters of small-sized LED irradiation chamber. Lect. Notes Netw. Syst. 569, 632–641 (2023) 12. Petrukhin, V., et al.: Modeling of the device operating principle for electrical stimulation of grafting establishment of woody plants. Lect. Notes Netw. Syst. 569, 667–673 (2023)

Investigation of Hydrodynamic Behaviour in Rectangular Sheet Shaped Membrane Using Computational Fluid Dynamics (CFD) Anirban Banik1(B) , Sushant Kumar Biswal2 , Tarun Kanti Bandyopadhyay3 , Vladimir Panchenko4,5 , Sunil Garhwal6 , and Anil Garhwal7 1 Department of Civil Engineering, National Institute of Technology Sikkim, Ravangla, South

Sikkim 737139, India [email protected] 2 Department of Civil Engineering, National Institute of Technology Agartala, Jirania, Tripura (W) 799046, India 3 Department of Chemical Engineering, National Institute of Technology Agartala, Jirania, Tripura (W) 799046, India 4 Russian University of Transport, Obraztsova St., Moscow 127994, Russia 5 Federal Scientific Agroengineering Center VIM, 1st Institutsky Passage 5, 109428 Moscow, Russia 6 Department of Civil Engineering, Maharishi Markandeshwar (Deemed to be University), Mullana, Ambala, Haryana 133203, India 7 School of Architecture and Planning, Banasthali Vidyapith, Tonk, Rajasthan 304022, India

Abstract. Current research investigates hydrodynamic behaviour in rectangular sheet shaped membrane implementing computational fluid dynamics approach. CFD mimics the flow phenomena in stationary and rotating module. The flow within the stationary and rotating membranes was computed using commercially available CFD software (ANSYS). It was well established that membrane can generate permeate flux of high quality but limited in its use due to its high fouling tendency. So, the effect of rotation in improving the antifouling property was also studied at 30 RPM. The results show good agreement with the experimental values. Keywords: Computational fluid dynamics · Membrane separation technique · Rotating membrane · Wastewater

1 Introduction Membrane filtration and separation processes have been critical in the industrial separation process for decades [1]. Various studies have been conducted to select the best method of membrane separation. The use of computational fluid dynamics (CFD) approach yields a wealth of information about the evolution of the process [2]. Various technological advancements in the field of membrane technology have simplified and expedited the process of selecting a membrane for a particular procedure. Hydrodynamics behaviour is vital in membrane separation and filtration techniques [3–5]. Flow © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 170–180, 2024. https://doi.org/10.1007/978-3-031-50158-6_18

Investigation of Hydrodynamic Behaviour in Rectangular

171

within Membrane separation is a process that combines free flow within the module with porous zone flow within the membrane bed. Fluid flow inside the module is simple to mimic whereas flow inside the porous zone is considered as the complex one to simulate which is achieved by hybrid form of Darcy law coupled with Navier-Stokes equation [6–8]. The approach of using Darcy law to predict solution for incompressible flow inside porous zone was found acceptable in the literature. The critical component is to ensure adequate continuity of flow field variables at the interface between laminar and porous flow regions [9, 10]. Numerous studies demonstrate the application of computational fluid dynamics (CFD) in membrane process optimization [7–13]. Gruber et al. investigates the ability of computational fluid dynamics to mimic the forward osmosis membrane operation. Even when high cross-flow velocity and slip velocity at the porous surface are taken into account, simulations show a non-negligible external concentration polarisation on the porous support [14]. Shirazian et al. employed CFD to predict hydrodynamics and mass transfer in membrane reactors implemented for ammonia separation from aqueous solutions. The system’s mass transfer and hydrodynamics were investigated by solving conservation equations. The findings of this study confirmed that feed and solvent velocities are the most significant parameter in ammonia removal [15]. Wang et al. simulated the zones of membrane filtration in submerged hollow fibre membrane using CFD. To account for the hydrodynamic behaviour of a full-scale submerged MBR, this porous media model was coupled with a 3D multiphase model [16]. The objective of the chapter is to employ computational fluid dynamics to simulate the flow inside the membrane bed to investigate the hydrodynamic behaviour in rectangular sheet membrane. Present study also investigates the effect of inlet velocity and membrane rotation. CFD predicted results are compared and evaluated against the experimental results.

2 Experimental Procedure An experimental study of the Cellulose acetate (CA) rectangular sheet shaped membrane was carried out in the laboratory to investigate the flow phenomena, permeate flux, and other properties of the membrane and to improve the quality of effluent from the rubber industry in Tripura, India. The setup of rectangular sheet shaped membrane includes neutralizing tank (used to monitor any undesirable variation in feed pH and temperature), feed tank, membrane module, and permeate tank. The feed stream has been allowed to flow with the assistance of a centrifugal pump. Rubber industry Effluents are collected from the Rubber Park located in TIDC Bodhjung nagar comlex. Complex is located in Bodhjung nagar village, which is about 12 km from the state capital Agartala, Tripura India. Characterization of raw effluent of rubber industry has been given in Table 1.

3 Computational Fluid Dynamics (CFD) Computational fluid dynamics (CFD) is a subfield of fluid mechanics that is used to forecast fluid flow and related problems by solving the governing equations for the flow phenomena [17–19]. The CFD technique aids in the collection of data regarding the flow phenomena occurring within the membrane module. CFD minimizes time and cost

172

A. Banik et al. Table 1. Characterization of raw effluent.

Sl. No

Parameters

Units

Value

1

pH

–

5.4

2

Total suspended solids

mg/l

3900

3

Total dissolved solids

mg/l

380

4

Sulphide

mg/l

16.5

5

Oil and grease

mg/l

11.5

6

BOD5

mg/l

724

of experimentation and data collection. The model of rectangular sheet membrane was developed implementing 3D hexahedral mesh using ANSYS. The flow in membrane bed is assumed to be laminar which was found from Reynolds number and it was found less than 2300. The flow through the membrane is governed by equations such as continuity, Darcy law, momentum, and solute transfer demonstrated by Eqs. (1)–(4), respectively [20].       d d d − → =0 (1) V i +j +k dx dy dz  μ − → P = − V (2) α   dP  2  ∇ ρuV = − (3) + μ∇ u dx     ∇ ρ V C = ρD∇ 2 C (4) 





In Eqs. (1)–(4), ρ, μ, V , and k, denotes the density, viscosity, velocity vector, and permeability, respectively. Diffusive coefficient and concentration are represented in Eq. 4 by D and C, respectively. 3.1 Assumptions While developing the theoretical and mathematical model describing effluent flow through membrane bed, the following assumptions and concepts are taken into account: I. II. III. IV.

Rubber industry effluent was considered as Newtonian fluid. Effluent was considered as incompressible and Isothermal. Membrane bed was considered isotropic and homogeneous porous zone. The under-relaxation factor was reduced to 0.7–0.3 or lower due to the necessity of the simulation work. V. For computational simplification, hexahedral grids are assumed. VI. Refined grids were used to resolve a large gradient of pressure and swirl velocities. VII. Developed model for simulation was limited to flow model only.

Investigation of Hydrodynamic Behaviour in Rectangular

173

3.2 Equations for Grid Discretization The governing equation was obtained by using the integral method. The fluid region of the concerned problem discretized into a set of control volume or cells. The transport equation for momentum, mass transfer or solute transfer, etc. has been applied to each cell and discretized for cell q given by Eq. (6); ∂ (ρφ)dV + (ρφu)dA = (∇φ)dA + Sφ dV (6) ∂t V

A

V

A

The general property of φ used for transport equation and each transport equation discretized into a system of the algebraic equation which is then solved numerically by rendering the solution domain of the problem using Eq. (7); t+t

t

− ρφp ρφp ρf φf uf Af = f Af (∇φ)f + Sφ V V + t faces

(7)

faces

Discretized equations need information at the centers of cell and faces. Required field data like velocity, properties of the material, etc. are stored at the centers of the cell. Face values are calibrated by interpolating the local and adjacent cell values. The accuracy of discretization depends upon the size of the stencil. The discretized equation expressed merely by using Eq. (8); (αp φp ) + (αnb φnb ) = bp (8) nb

3.3 Boundary Conditions The above governing equation of the problem has been solved subject to the below mentioned boundary conditions [21]: (a) (b) (c) (d)

Inlet of the membrane was considered to be mass inlet. Outlet of the membrane was assumed to be pressure outlet. Membrane was assumed to porous zone. At the membrane module’s wall, no slip condition was taken into account.

3.4 Convergence and Test of Grid Independency Except for the transport equation, which was set to 10–3 , the default convergence criteria for all equations were set to 10–5 . A computational domain was used to calibrate the findings of fully developed flow that might be acquired for membrane. Study illustrates that pronounced results get affected by the mesh size. To check if the chosen mesh/grid resolution was adequate to produce results with a low error percentage, the mesh/grid resolution was gradually increased and decreased by 50%. A reduction in mesh resolution was shown to contribute 8–15% of the inaccuracy in the pressure profile, whereas an increase of 50% in mesh resolution reduced the error to 1–5%. The grid size type

174

A. Banik et al.

is illustrated using Table 2. Grid independency test was reported in Table 3. The fine grid of size 192000 was determined to be the best grid size for carrying out the simulation procedure, since increasing the grid size results in no significant change in the pressure profile of the membrane. Based on the findings, it is possible to infer that the existing mesh/grid resolution is sufficient for producing grid independent solutions for the suggested membrane model. Table 2. Grid size type Sl. No

Grid type

Grid size type

Grid size

No. of nodes

No. of faces

1

Hexahedral

Coarse

98304

105633

302080

2

Hexahedral

Fine

192000

203401

587200

3

Hexahedral

Finer

375000

392751

1E+06

Table 3. Grid independency test Sl. No

Grid type

Grid size

Time (s)

Expt. (kPa)

CFD (kPa)

Error (%)

1

Hexahedral

98304

80

2

Hexahedral

192000

110

46

55.6

20.8

46

43.4

5.6

3

Hexahedral

375000

156

46

43.4

5.6

3.5 Advantages and Disadvantages of CFD CFD has the following advantages: it provides insight into the system that would be impossible to obtain through experimentation, it predicts efficient and optimal design, it minimises experimentation costs, it can simulate ideal and real conditions, and it can investigate a large number of locations. The following are the disadvantages of CFD: The accuracy and efficiency of the CFD model are dependent on the physical model of the real world; solving the governing equations numerically via a computer results in the accumulation of numerical errors such as round-off and truncation errors. Additionally, the accuracy of the model is influenced by the physical model’s initial boundary condition.

4 Results and Discussions 4.1 CFD Analysis The static pressure (Pa) contour for a rectangular sheet membrane used to improve the effluent quality of the rubber industry is shown in Fig. 1. The plot demonstrates that pressure steadily declines over the membrane bed, which is caused by the membrane’s

Investigation of Hydrodynamic Behaviour in Rectangular

175

high resistance. Kinetic energy heads are lost along the membrane because to its high resistance, and this loss of kinetic energy causes a pressure decrease across the membrane. Figures 2 and 3 depict a rectangular sheet membrane’s wall shear stress (Pa) and shear strain rate (sec−1 ). The figures demonstrate that wall shear stress and shear strain are high at the membrane’s wall and pore wall because of the high skin friction.

Fig. 1. Contour of Static pressure (Pa)

Fig. 2. Contour of wall shear stress (Pa)

Fig. 3. Contour of Shear strain Rate (sec−1 )

176

A. Banik et al.

4.2 Effect of Inlet Velocity Figure 4a–d illustrates static pressure (Pa) contour for effluent flow inside the membrane at various inlet velocities. According to the plot, as the feed stream velocity at the membrane module’s entrance increases, so does the static pressure on the membrane module. This increase in speed is depleting kinetic energy. The high membrane resistance that obstructs the flow and manifests as a type of pressure energy across the membrane is to blame for this. 4.3 Effect of Membrane Rotation The main disadvantage associated with the membrane separation and filtration is fouling. From the study of the static membrane, it is found that the membrane undergoes rapid fouling due to the blockage in the membrane pores and formation of cake layer over the membrane bed resulted from the deposition of macromolecules and impurities present in the feed stream. The rotational rectangular sheet membrane is used to enhance the permeate quality, filtration time and antifouling property of the membrane. Figure 5 illustrates static pressure (Pa) contour of rotating rectangular sheet membrane with a rotating speed of 30 RPM. From this figure, it found that the rotational rectangular sheet membrane has more inlet pressure and pressure drop compared to the static membrane bed. This is due to its high resistance and high shear which acts on the membrane bed. The reason of generation of high shear is due to the rotation of the membrane bed. Red and blue colour in the plot illustrates the region of high and low pressure respectively. The plot also shows the wave formation over the membrane bed due to its rotation. The rotation of membrane minimizes the fouling tendency by keeping the macromolecules and impurities in the suspended forms. This prevents the formation of the cake layer over the membrane. Figure 6 demonstrates the velocity (m/sec) contour of rotating rectangular sheet membrane. The flow through the permeable layer, over the membrane bed, and in the bulk feed is depicted on the plot. Due to the large adhesive force between the wall and fluid particles compared to the cohesive force between the liquid molecules, it was discovered that flow tends to cease near the wall. However, rotation in the membrane module forces the feed to flow near the wall. This is due to the centrifugal force which pushes the fluid towards the wall. 4.4 Validation The graphical illustration of pressure drop (kPa) and inlet velocity (m/sec) of rectangular sheet membrane is shown in Fig. 7. Plot illustrated that increased inlet velocity influences the fluid momentum as the momentum is a function of velocity. The fluid’s kinetic energy head is turned into a kind of pressure head when the inlet velocity increases, resulting in a greater pressure drop across the membrane. Figure 7 also illustrates the CFD pronounced values follows the experimental value with higher accuracy.

Investigation of Hydrodynamic Behaviour in Rectangular

177

(a)

(b)

(c)

(d) Fig. 4. Static pressure (Pa) contour for rectangular sheet membrane with velocity at inlet as: (a) 17.9 cm/sec (b) 30.14 cm/sec (c) 58.18 cm/sec and (d) 82.82 cm/sec

178

A. Banik et al.

Fig. 5. Contour of static pressure (Pa) for rectangular membrane with a rotating speed of 30 RPM.

Fig. 6. Contour of velocity (m/sec) for rectangular sheet membrane with a rotating speed of 30 RPM.

Fig. 7. Plot of the membrane’s inlet velocity versus pressure drop

Investigation of Hydrodynamic Behaviour in Rectangular

179

5 Conclusions, Limitations, and Future Research Scope In present study, fluid flow in rectangular sheet shaped membrane was modelled using ANSYS. To pronounce precise solution hexahedral mesh was implemented for meshing purpose. Based on the outcome of grid independency test, grid size of 192000 was chosen for simulation purpose. Moreover, it was also found that any additional refinement of the grid size has no effect on the pressure profile of the rectangular sheet membrane. According to the study, rotating membranes outperform static membranes in terms of filtering life. The predicted data of the CFD are compared to the experimental one, which show excellent agreement with one another, and the error percentage typically ranges between 1 and 5%. CFD analysis provides a clear picture about the distribution of pressure, wall shear stress, velocity, and shear strain rate in membrane module. Experimental run for rotating membrane was not conducted which is one of the major limitations of the study. In future, experimental run will be conducted for better validation of the rotating membrane. Moreover, optimization of rotating speed of the membrane will also be investigated in future. The study’s findings can be used to develop a low-cost membrane separation technique for treating the effluent generated from rubber industry.

References 1. Stopford, P.J.: Recent applications of CFD modelling in the power generation and combustion industries. Appl. Math. Model. 26, 351–374 (2002). https://doi.org/10.1016/S0307-904 X(01)00066-X 2. Kim, S.E., Boysan, F.: Application of CFD to environmental flows. J. Wind Eng. Ind. Aerodyn. 81, 145–158 (1999). https://doi.org/10.1016/S0167-6105(99)00013-6 3. Vasant, P., Zelinka, I., Weber, G.-W.: Intelligent Computing and Optimization-Proceedings of the 3rd International Conference on Intelligent Computing and Optimization 2020 (ICO 2020) (2021). https://doi.org/10.1007/978-3-030-68154-8 4. Vasant, P., Zelinka, I., Weber, G.-W. (eds.): ICO 2019. AISC, vol. 1072. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-33585-4 5. Vasant, P., Zelinka, I., Weber, G.-W. (eds.): ICO 2018. AISC, vol. 866. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-00979-3 6. Karode, S.K.: Laminar flow in channels with porous walls, revisited. J. Membr. Sci. 191, 237–241 (2001). https://doi.org/10.1016/S0376-7388(01)00546-4 7. Pak, A., Mohammadi, T., Hosseinalipour, S.M., Allahdini, V.: CFD modeling of porous membranes. Desalination 222, 482–488 (2008). https://doi.org/10.1016/j.desal.2007.01.152 8. Berman, A.S.: Laminar flow in channels with porous walls. J. Appl. Phys. 24, 1232–1235 (1953). https://doi.org/10.1063/1.1721476 9. Weber, G.W., Taylan, P., Alparslan-Gök, S.Z., Özöˇgür-Akyüz, S., Akteke-Öztürk, B.: Optimization of gene-environment networks in the presence of errors and uncertainty with Chebychev approximation. TOP 16, 284–318 (2008). https://doi.org/10.1007/s11750-0080052-5 10. Ozogur-Akyuz, S., Weber, G.W.: On numerical optimization theory of infinite kernel learning. J. Global Optim. 48, 215–239 (2010). https://doi.org/10.1007/s10898-009-9488-x 11. Ghadiri, M., Asadollahzadeh, M., Hemmati, A.: CFD simulation for separation of ion from wastewater in a membrane contactor. J. Water Process Eng. 6, 144–150 (2015). https://doi. org/10.1016/j.jwpe.2015.04.002

180

A. Banik et al.

12. Rahimi, M., Madaeni, S.S., Abolhasani, M., Alsairafi, A.A.: CFD and experimental studies of fouling of a microfiltration membrane. Chem. Eng. Process. 48, 1405–1413 (2009). https:// doi.org/10.1016/j.cep.2009.07.008 13. Rezakazemi, M.: CFD simulation of seawater purification using direct contact membrane desalination (DCMD) system. Desalination 443, 323–332 (2018). https://doi.org/10.1016/j. desal.2017.12.048 14. Gruber, M.F., Johnson, C.J., Tang, C.Y., Jensen, M.H., Yde, L., Hélix-Nielsen, C.: Computational fluid dynamics simulations of flow and concentration polarization in forward osmosis membrane systems. J. Membr. Sci. 379, 488–495 (2011). https://doi.org/10.1016/j.memsci. 2011.06.022 15. Shirazian, S., Rezakazemi, M., Marjani, A., Moradi, S.: Hydrodynamics and mass transfer simulation of wastewater treatment in membrane reactors. Desalination 286, 290–295 (2012). https://doi.org/10.1016/j.desal.2011.11.039 16. Wang, Y., Brannock, M., Cox, S., Leslie, G.: CFD simulations of membrane filtration zone in a submerged hollow fibre membrane bioreactor using a porous media approach. J. Membr. Sci. 363, 57–66 (2010). https://doi.org/10.1016/j.memsci.2010.07.008 17. Banik, A., Bandyopadhyay, T.K., Biswal, S.K.: Computational fluid dynamics (CFD) simulation of cross-flow mode operation of membrane for downstream processing. In: Recent Patents on Biotechnology, vol. 13 (2019). https://doi.org/10.2174/187220831266618092416 0017 18. Banik, A., Bandyopadhyay, T.K., Biswal, S.K., Majumder, M.: Prediction of maximum efficiency of vertical helical coil membrane using group method of data handling (GMDH) algorithm. In: Vasant, P., Zelinka, I., Weber, G.-W. (eds.) ICO 2019. AISC, vol. 1072, pp. 489–500. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-33585-4_48 19. Debnath, A., Banik, A., Bandyopadhyay, T.K., Saha, A.K.: CFD and optimization study of frictional pressure drop through bends. In: Recent Patents on Biotechnology, vol. 13 (2019). https://doi.org/10.2174/1872208312666180820153706 20. Banik, A., Biswal, S.K., Bandyopadhyay, T.K.: Predicting the optimum operating parameters and hydrodynamic behavior of rectangular sheet membrane using response surface methodology coupled with computational fluid dynamics. Chem. Pap. 74(9), 2977–2990 (2020). https://doi.org/10.1007/s11696-020-01136-y 21. Banik, A., Bandyopadhyay, T.K., Biswal, S.K.: Computational fluid dynamics simulation of disc membrane used for improving the quality of effluent produced by the rubber industry. Int. J. Fluid Mech. Res. 44, 499–512 (2017). https://doi.org/10.1615/InterJFluidMechRes.201 7018630

A Review on the Impacts of Social Media on the Mental Health Md. Abu Bakar Siddiq Tapu1 , Rashik Shahriar Akash1 , Hafiz Al Fahim1 , Tanin Mohammad Jarin1 , Touhid Bhuiyan1 , Ahmed Wasif Reza2(B) , and Mohammad Shamsul Arefin1,3(B) 1 Department of Computer Science and Engineering, Daffodil International University, Birulia,

Bangladesh {abu15-3782,rashik15-3825,hafiz15-3781,tanin15-3812}@diu.edu.bd, [email protected] 2 Department of Computer Science and Engineering, East West University, Dhaka, Bangladesh [email protected] 3 Department of Computer Science and Engineering, Chittagong University of Engineering and Technology, Chattogram, Bangladesh

Abstract. There are numerous effects of social media use on people’s daily lives. Every day we are connected a lot of time with social media. As a result, our brains become unbalanced, and we feel a lot of illness in our bodies. As a result, the primary objective of this analysis is to provide information on how social media affects its users. Nineteen studies were included in this paper regarding the main purpose. We categorized the papers into four types, Status and Post based research, Research paper and Database analysis-based research, Survey based research, and disease-based research. After a comprehensive analysis of the available literature, we have concluded that utilizing the social side has an effect on our mental health and actions. The results highlighted a number of critical aspects, such as the varied approaches to determining the influence of social media, the limitations of the studies, and our thoughts on what should be enhanced. Keywords: Mental health · Database analysis · Disease-based research · Privacy concern · COVID-19

1 Introduction Advancement and sophistication of technology, social communication media has become a ubiquitous part of everyday life. We are connected with the media either by the need of time or unknowingly. According to experts, excessive social media use becomes addictive. As a result, our brains become unbalanced in the absence of social media. People become dependent on social media to feel ‘normal’. And behavior controlled by addiction is never healthy. Put another way, the mental health of users is highly influenced by social media.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 181–195, 2024. https://doi.org/10.1007/978-3-031-50158-6_19

182

Md. A. B. S. Tapu et al.

We analyzed a number of papers in which researchers have extensively discussed the potential relationship between social media and psychological negative effects. Politically provocative posts seen on social media, anger-frustration, inflammatory, offensive statements, etc. from various acquaintances also affect your mental health. Which will only incite anger and resentment in your mind [2, 5]. Research has revealed that prolonged use of social media is associated with negative emotions like depression, isolation, and violence [14]. A common keyword found in all the papers we selected was ‘concern’. In the social media world, this concern is reflected in various aspects including security [1], mental fatigue, failure, self-risk, misuse of time, etc. In addition, a large part of our knowledge gained was related to the mental health of young adolescents [2, 6, 7, 9, 15]. Social media has been found in multiple studies to have a significant impact on how people organize their time and lives. However, this issue is very worrisome and challenging. Experts point out many factors as reasons for this, such as—misuse of technology, mismanagement of time, downloading unnecessary apps, following trends, etc. However, the papers show that data has been collected directly and indirectly from people of different levels and professions and implemented through different algorithms for outcome analysis. Considering all aspects, we can conclude that social media or social networking is deeply related to the mental health of individuals. Which is a significant issue in today’s society.

2 Methodology To summarize the most up-to-date research methodology practices in a substantive field or issue, a methodology review is a subset of systematic secondary research. In this paper, we included nineteen papers from various sites and we included different types of papers like journals, conferences, etc. This paper is a review of the details of these Nineteen papers. 2.1 Phase 1: Planning This is the section that describes how we selected the papers from various places. The searching method was, we first identified the topic, then We learned about the topic or social media impact then we searched on various platforms like Google Scholar, Refseek, Microsoft Academic, and IEEE access about the topic. Then we selected the papers which are related to our purpose. We mostly try to collect the last 5 years’ papers and some of them fulfilled our purpose. The search terms we used to find the existing papers include “Social Media Use and Mental Health” “The influence of social media on depression, anxiety”, “is social media bad for mental health and well-being”, “Effects of Social Media on Adolescents” etc. 2.2 Phase 2: Conducting This section described the way how the existing papers were checked. We have taken enough time to read the papers to get the best outcome from the papers. We read the

A Review on the Impacts of Social Media on the Mental Health

183

title and abstract first to know whether the paper actually fulfilled our purpose or not. Then if we thought the paper was right for our research, we read the conclusion then. Then we read the methodologies, results, and discussion of every paper to get the proper information. 2.3 Phase 3: Reporting After choosing and reviewing the papers, the authors categorized the papers into four types, Status and Post based research, Research paper and Database analysis based research, Survey based research, and disease-based research.

3 Paper Collection Over time, the technology of social media has also improved, as technology has improved, it has attracted people in different ways to use social media or it has been seen that over time, people have been attracted to use social media for different reasons. It has been seen that various physical and mental changes have occurred among users due to the use of social media over time. Based on these, it is seen that there is some difference between the research paper data of one year and the research paper data of the next or previous year. Since there are many research papers related to our work, we have started from the year 2016 in the case of paper selection because it has been seen in the previous data that there are no significant physical or mental changes or problems among social media users. Moreover, we have selected only those papers which are related to our research. We have tried to do our research based on the analysis of four types of categories: Status and Post based, research paper and database-based, survey-based, and disease-based. So based on these four types, among the papers we found, some years have fewer papers of that type. In this case, In Table 1, we have merged some years like 2016–2017, and 2020–2021. And based on these, we have sorted the information into 5 columns in the form of a paper collection table, it includes the topic type and based on that the information of the paper available in the years 2016–2017, 2018, 2019, 2020–2021 (Fig. 1). Table 1. Evolutions of paper collection on impact analysis of social media on the mental health of the users Topic types

2016–2017

Status and post based

2018

2019

2020–2021

Naslund, J.A [1]

Vogel, E.A [9]

Thorstad [12] Tadesse [5] Research paper and database Baker, D.A. [15] analysis based Wongkoblap [16]

Chancellor [2]

Keles [6]

(continued)

184

Md. A. B. S. Tapu et al. Table 1. (continued)

Topic types

2016–2017

2018

Survey based

Pontes [17]

Berryman [3] Scott [8]

2019

Andreassen [13]

O’reilly [7]

Cao, X., Khan [10]

Lau, W.W [19]

Rahman [18]

Azizi [11]

Disease based

2020–2021

Zhao [4] Ahmad, A. R [14]

The below graph represents that most of the papers we analyzed are based on questionnaires and availability of these types of papers is high but we have tried our best to analyze every type of paper that is relevant to our research type.

Fig. 1. Paper collection described in pie format

4 Detailed Review of Papers This section focuses on the issue addressed, source of data, usefulness and assessment of each work. Table 2 illustrates the Status and Post based analysis where writers have offered their ideas on effect analysis of social media on the mental health of the users. In [1], They utilized a Qualtrics poll to get the information from 90 respondents. Since May 2016, they have independently studied how individuals with mental illnesses use the social media site Twitter. They learned about the user’s condition through this, including reluctance, fear, interpersonal relationships, and changes in other people’s mental states.

A Review on the Impacts of Social Media on the Mental Health

185

Table 2. Status and post based research Paper title

Problem addressed

Source of data

Evaluation

Risks to privacy with use of social media [1]

They mainly tried to identify the interaction of users’ mental health with personal security concerns on social media

They have studied social media site Twitter since May 2016

They discovered that around a third of the people in their study admitted to having a mental disorder. They experience mental weariness because they are unaware of some technical problems and privacy dangers

Effects of social Higher e-cigarette Ads and social media media on adolescents usage intentions were content allegedly [9] observed, and In such written by teenagers a short amount of time, individuals who were exposed to e-cigarette material on social media increased their usage

Highlights important details, such as how teens’ frequent usage of social media and their brief exposure to content about e-cigarettes on such platforms influence their use of those platforms. It also suggests that government action be done to prohibit promoted e-cigarettes on social media

Predicting future mental illness [12]

Their findings showed that even in patients who did not have a history of mental illness, the impacts of anxiety and disease spread gradually. It poses a serious mental health challenge for the next generation

Based on the language and signals used by social media users, three topics relating to mental illness—Clinical Subreddits, Nonclinical Subreddits, and Future Prediction—were highlighted

Gathered from Reddit media about users’ words and actions

(continued)

186

Md. A. B. S. Tapu et al. Table 2. (continued)

Paper title

Problem addressed

Source of data

Detection of depression [5]

This paper attempts Reddit media platform to identify the main users causes of emotional exhaustion and depression among online users based on the posts

Evaluation Define a deeper relationship between depression and how one speaks They also identified some vocabulary words that are associated with users’ feelings of depression, anxiety, anger and suicidal tendencies which may badly affect both in the present and in the future

Study of these authors [9] explained the impact of short exposure to social media material, including advertisements and posts reportedly made by teenagers, and the correlation between daily social media usage and attitudes and intentions toward ecigarettes and desire to use e-cigarettes among adolescents. From Reddit media, they gathered indicators of users’ language and behavior. Based on these posts, they concentrated on mental illness and nonclinical subreddits and categorized clinical subreddits as well. They analyzed the data using the clustering technique. Based on theories that supported the patterns of mental disease, they predicted the future [12]. They looked at the posts made by roughly 100 university students. For optimal outcomes on Reddit, they further use Natural Language Processing (NLP) methods and machine learning algorithms, they classified user-generated regular posts and postings indicative of depression separately. They were able to predict 93% of F1 results after putting each phase into practice, beginning with data pre-processing (via NPL). They succeeded in establishing a more direct link between language use and depression [5]. Table 3 represents the Research paper and Database analysis based research. Where [2] in order to forecast users’ moods, behaviors, and attitudes, they gathered information using machine learning techniques, natural language processing (NLP), and human computer interface (HCI). They have classified it depending on the ethical difficulties that the consumers encountered. They divide “ethical tension” into the three sections listed below: Participants and research monitoring are listed in I. Methods, validity, and interpretability are listed in II. 3. Repercussions for Stakeholders. In [6], This investigation suggests that a number of variables interact to influence the extent to which adolescents’ use of social media is associated with increased rates of depression, anxiety, and other forms of psychological distress. To reduce the prevalence

A Review on the Impacts of Social Media on the Mental Health

187

of anxiety disorders among teenagers at high risk, we must address not just the symptoms of depression, but also associated outcomes such as anxiety and psychological distress. Table 3. Research paper and database analysis based Paper title

Problem addressed

Source of data

A taxonomy of ethical tensions [2]

Based on the Clinical data behavior, mood, and other characteristics of social media users, they produced a forecast of mental health here. The propensity for people to hurt themselves as a result of being persuaded by these social media sites has been the subject of substantial research

Evaluation They basically provide a rough taxonomy of issues with social media data-based algorithmic mental health status prediction. These fundamental studies on depression have characterized new psychiatric disorders and their relationships with contemporary social media

A systematic review: Examined how teen the influence [6] social media use affects their level of anxiety, despair, and discomfort

Five databases of other papers. Analyzed (Medline, Embase, PsychINFO, CINAHL and SSCI)

Add to the current literature by filling in the blanks and drawing attention to the significance of the phenomena of the mental health effect of social media usage among teenagers

The relationship between online [15]

Five databases were identified: PsycINFO, Web of Science, CINAHL, MEDLINE, and EMBASE

Several different psychological, social, behavioral, and individual variables may interact to produce a complicated relationship between online social networking and depressed symptoms

Determine whether there is a correlation between using social networking sites and experiencing depressive symptoms

(continued)

188

Md. A. B. S. Tapu et al. Table 3. (continued)

Paper title

Problem addressed

Source of data

Researching mental health [16]

Examine the 2010–2017 medical applicability and and computer science constraints of journals cutting-edge techniques employed by scientists to conduct mental health predictive analytics

Evaluation Contributed to the existing literature in a way of providing considerable evidence for the mental health impact of social media use by focusing on not only the symptoms of depression but also other related outcomes

The authors chose [15] Five databases—PsycINFO, Status and Web of Science, CINAHL, MEDLINE, and EMBASE—for this study. Protocol for Improved Reporting of Epidemiologic Observational Research The quality of cross-sectional research was evaluated using a statement approach. In total, 35 044 people took part in the studies that were examined. Studies were discovered in 14 different countries. The participants ranged in age from 15 to 88. Only three of the remaining 27 studies were longitudinal or had longitudinal components; the rest all utilized cross-sectional designs. Only correlation analysis was performed in seven studies, whereas regression analysis was used in 23. Although some research concentrated on certain factors like the amount of time spent social networking, All the investigations focused on the real links between depressive symptoms and use of social networking sites. Cross-sectional studies’ reporting quality ranges from 48 to 93%, according to the STROBE rating. In article number [16], they chose 48 publications for review out of a total of 5386, and they found that these studies use two different methods to gather data. They coded the data using articles’ data in accordance with essential characteristics, data collection methods, and data preprocessing, and analyzed the applicability and limitations of cutting-edge methods used by scientists to do predictive analytics on mental health. The sample data used in the supervised learning techniques examined in this study contains labeled inputs as well as outputs. These methods teach the model how to foresee and produce forecasts for unlabeled inputs from various sources. The results from the questionnaires are shown in Table 4. Scientists that wrote this study [3]. The sample included 471 college freshmen located in the Southeast of the United States. The results demonstrated that social media traits did not effectively predict bad outcomes. The risk of feeling alone and having suicidal thoughts was predicted by ambiguous booking with few exceptions. Neither how much time was spent online nor the weight of social media had a role in determining the outcome. This work [7] provides details that they incorporated three concepts, each of which is reflected in this article because they address the goals and focus of the study’s subject and show the range of participants’ social media views on mental health and wellness. This paper [8] provides details that they used six self-report measures to gather information from over 12,000 teenagers. Teenagers divulge details on their daily social

A Review on the Impacts of Social Media on the Mental Health

189

Table 4. Survey based research Paper title

Problem addressed

Source of data

Evolution

Social media use [3] A person’s mental or The sample collected physical health issue from 471 university may be traced back to students their time spent on social media. It’s possible that’s not the case

Only social media use is not blamed for mental health conditions rather than social media, other incidents are also related to mental health problems

Is social media bad [7]

The three problems highlighted mainly (mood and anxiety disorder, cyberbullying, use of social media addiction) gives an idea and helps future researchers to find out what are the causes of mental health problems because of social media use

Addressed the 11–18 years old perception among school students from teenagers that their two cities psychological well-being is being compromised by their social media use

Social media use [8] Dependencies on Data from nearly online platforms in all twelve thousand aspects affect adolescents teenagers’ health and sleep patterns

The stimulators of social media [10]

Mainly highlighted the extent to which students are exposed to various negativity psychologically on online platforms, which cause serious damage to their mental health

Teenagers’ interactions with online platforms and sleep have many negative side effects that could cause them great harm in the future. Through this survey, they have been able to provide

14–22 college students Their methodologies show that SMA functions as a particular class of stimulant that alters users’ psychological characteristics, including their levels of rage, agitation, and irritability (continued)

media and sleep routines. They looked into social media use and daily sleep characteristics using binomial logistic regression. A survey on social media usage by gender revealed that boys utilized it more frequently than girls.

190

Md. A. B. S. Tapu et al. Table 4. (continued)

Paper title

Problem addressed

Source of data

Evolution

The relationship between social [11]

Students’ participation in extracurricular activities and academic achievement are adversely affected by their use of social media

Survey of 360 medical Among the pupils science students questioned, there was a mild trend toward social media addiction. Their compulsive behavior, however, had a detrimental effect on their academic performance

Investigating the differential [17]

Examining the interactions between SNS addiction and IGD, as well as how they contribute to the development of psychiatric distress

Sample collected from They discovered that 509 students important demographic factors can contribute significantly to the understanding of SNS addiction and IGD

FACEBOOK USE [18]

Intended to examine how Facebook addiction and usage affect mental health

209 University students

The findings revealed that women used Facebook more frequently than men

Andreassen [13]

Talk about the physical symptoms that might arise from an addiction to social media and video gaming in adults

24,000 users, mainly 36-year-old users on average

Explained that addiction to social media and video games has been linked to a variety of underlying mental problems, including ADHD, OCD, anxiety, and depression

Effects of social [19] How usage of social From 348 university media or multitasking students of social media affect academic performance

Those who have used social media for study purposes, have no significant impact on their academic performance rather than using social media for nonacademic purposes. It has a negative impact on their life

Using a stratified sampling technique, data from college students between the ages of 14 and 22 were gathered. They employed structural equation modeling to test their measurement and structural model. They investigated the internal connections between

A Review on the Impacts of Social Media on the Mental Health

191

SMA and students’ psychological suffering as a result, and they confirmed that their findings were compatible with those of earlier research [10]. For this research [11], 360 Iranian students pursuing medical science participated in a stratified random sampling survey. To determine the degree of these respondents’ addiction, the Bergen Social Media Addiction test was employed. They also looked at the information related to the pupils’ performance and academic progress. In this study [17], they get information from 509 students. The sample’s median age was 13.02 years (SD: 1.64), and the gender distribution was roughly equal, with 53.5% of men (n = 265). Age, gender, and relationship status-related demographic data were gathered. By asking participants how much time they typically spend online each week, information about SNS usage was gathered. The sample was given nine questions, the total value of which ranged from 9 to 45, with each question’s value falling between 1 and 5. The more positive the value, the more GD is present. Credibility score was high ( = 0.87). In this research [18], through the questionnaires, they collect data from 209 university students. In questionnaires or data analysis they used these two: Bergen Facebook Addiction Scale, BFAS (Andreassen et al., 2012) and the Bangla (Ahmed et al., 2018) Mental Health Inventory(MHI)—18 (Veit and Ware, 1983) methods to compile a report on respondents’ Facebook addiction and mental health. In the BFAS method, higher scores indicate higher addiction to Facebook. Their MHI methods represent if there is a higher score then there is higher mental health. However, their result found that women spend more time on Facebook than men. As they spend more time on Facebook, users cannot allocate proper time for daily work and social relations etc. Further they faced physical or mental problems like anxiety, depression etc. Their extensive data [13] collecting sample size (about 24,000). Nevertheless, the key respondents’ average age was 36. They gathered user information and looked at their histories of social media and video game addiction through an online cross-sectional survey. Additionally, users provided self-reports of any current psychological conditions. They looked into the addictiveness of social media and video games using regression analysis techniques. Table 5 represents the Disease based analysis. The findings [4] of this study suggested that Poorer mental health was associated with more time spent on social media. Participants with high (but not low) levels of the disaster stressor had a positive association between increased exposure to catastrophe news on social media and increased depression. Route analysis also shown that negative emotions moderate the connection between social media usage and psychological well-being. This work [14] analysis, effects of online panic during the COVID-19 To learn more about the pandemic, researchers in Iraqi Kurdistan developed and administered an online survey to a representative sample of 516 users of social media in that region. Data from this research were analyzed using content analysis. SPSS study of the data supports this conclusion.

192

Md. A. B. S. Tapu et al. Table 5. Disease based research

Paper title

Problem addressed

Source of data

Evaluation

Social media use [4]

Aiming to learn and bring attention to the fact that prolonged exposure to catastrophe coverage on social media may lead to psychological distress

512 user data

The results indicate that exorbitant exposure to catastrophe news on social media has an effect on mental health and that future interventions aimed at enhancing it should take both disaster stressor and adverse effects into account

The impact of social [14]

Self-reported mental health and the propagation of COVID-19 panic are influenced by social media use

516 social media users were sampled

A self-reported higher frequency of social media usage was associated with a higher likelihood of COVID-19 fear

5 Discussion In this paper, we have described nineteen papers about the “Impact Analysis of Social Media on the Mental Health of the Users “. We reviewed all the factors and important points of all nineteen papers and after reviewing and understanding the papers, we have observed that there is a big impact of using the social side on our mental health but mental health problems are not caused by social media usage alone. From all the research paper analysis we notice that Higher levels of social media usage were related to worse psychological wellbeing. There is a vast variety of social media material that teenagers may encounter, which is very harmful to their mental health. Most of the papers did the analysis based on surveys that are open to self-identifying biases. We can easily measure the impact of the data analysis and survey which included the results of social media on the mental health of the users.

6 Conclusion It is worth noting that social media is inextricably intertwined with individual and social life. There is no doubt that social media has brought people from different parts of the world much closer to each other. However, according to the topic of our discussion, the message that these studies give us is that Directly and indirectly, excessive social media usage contributes to the development of a wide range of mental health issues. At least for adolescents, it is threatening. Influenced by these, they do not hesitate to indulge in various suicidal thoughts. However, the issues that could not be clearly described in the studies are the relationship between the use of social media and maintaining

A Review on the Impacts of Social Media on the Mental Health

193

time management. In addition, if more information from direct sources (such as social isolation) could have been added to the data collection, the outcome would have been more accurate. However, ultimately as important as a healthy lifestyle is, it is impossible to completely abstain from all social platforms. So, the benefits of social media should be taken by controlling its negative side effects. And in this way, it will be possible for a person to remain mentally healthy even if he is connected to social media platforms.

References 1. Naslund, J.A., Aschbrenner, K.A.: Risks to privacy with use of social media: understanding the views of social media users with serious mental illness. Psych. Serv. 70(7), 561–568 (2019) 2. Chancellor, S., Birnbaum, M.L., Caine, E.D., Silenzio, V.M., De Choudhury, M.: A taxonomy of ethical tensions in inferring mental health states from social media. In: Proceedings of the Conference on Fairness, Accountability, and Transparency, pp. 79–88 (2019) 3. Berryman, C., Ferguson, C.J., Negy, C.: Social media use and mental health among young adults. Psych. Quart. 89(2), 307–314 (2018) 4. Zhao, N., Zhou, G.: Social media use and mental health during the COVID-19 pandemic: moderator role of disaster stressor and mediator role of negative affect. Appl. Psychol. Health Well-Being 12(4), 1019–1038 (2020) 5. Tadesse, M.M., Lin, H., Xu, B., Yang, L.: Detection of depression-related posts in reddit social media forum. IEEE Access 7, 44883–44893 (2019) 6. Keles, B., McCrae, N., Grealish, A.: A systematic review: the influence of social media on depression, anxiety and psychological distress in adolescents. Int. J. Adolesc. Youth 25(1), 79–93 (2020) 7. O’reilly, M., Dogra, N., Whiteman, N., Hughes, J., Eruyar, S., Reilly, P.: Is social media bad for mental health and wellbeing? Exploring the perspectives of adolescents. Clin. Child Psychol. Psych. 23(4), 601–613 (2018) 8. Scott, H., Biello, S.M., Woods, H.C.: Social media use and adolescent sleep patterns: crosssectional findings from the UK millennium cohort study. BMJ Open 9(9), e031161 (2019) 9. Vogel, E.A., et al.: Effects of social media on adolescents’ willingness and intention to use e-cigarettes: an experimental investigation. Nicotine Tobacco Res. 23(4), 694–701 (2021) 10. Cao, X., Khan, A.N., Zaigham, G.H., Khan, N.A.: The stimulators of social media fatigue among students: role of moral disengagement. J. Educ. Comput. Res. 57(5), 1083–1107 (2019) 11. Azizi, S.M., Soroush, A., Khatony, A.: The relationship between social networking addiction and academic performance in Iranian students of medical sciences: a cross-sectional study. BMC Psychol. 7(1), 1–8 (2019) 12. Thorstad, R., Wolff, P.: Predicting future mental illness from social media: a big-data approach. Behav Res Methods 51(4), 1586–1600 (2019) 13. Andreassen, C.S., et al.: The relationship between addictive use of social media and video games and symptoms of psychiatric disorders: a large-scale cross-sectional study. Psychol. Addict. Behav. 30(2), 252 (2016) 14. Ahmad, A.R., Murad, H.R.: The impact of social media on panic during the COVID-19 pandemic in Iraqi Kurdistan: online questionnaire study. J. Med. Internet Res. 22(5), e19556 (2020) 15. Baker, D.A., Algorta, G.P.: The relationship between online social networking and depression: a systematic review of quantitative studies. Cyberpsychol. Behav. Soc. Netw. 19(11), 638–648 (2016)

194

Md. A. B. S. Tapu et al.

16. Wongkoblap, A., Vadillo, M.A., Curcin, V.: Researching mental health disorders in the era of social media: systematic review. J. Med. Internet Res. 19(6), e7215 (2017) 17. Pontes, H.M.: Investigating the differential effects of social networking site addiction and Internet gaming disorder on psychological health. J. Behav. Addict. 6(4), 601–610 (2017) 18. Rahman, M., Ahmed, O.: Facebook use, facebook addiction, and mental health of chittagong university students. Bulg. J. Sci. Educ. Policy 12(2), 14 (2018) 19. Lau, W.W.: Effects of social media usage and social media multitasking on the academic performance of university students. Comput. Hum. Behav. 68, 286–291 (2017) 20. Stavridou, A., et al.: Psychosocial consequences of COVID-19 in children, adolescents and young adults: a systematic review. Psych. Clin. Neurosci. 74(11), 615 (2020) 21. Naslund, J.A., Bondre, A., Torous, J., Aschbrenner, K.A.: Social media and mental health: benefits, risks, and opportunities for research and practice. J. Technol. Behav. Sci. 5, 245–257 (2020) 22. Fernandes, B., et al.: Internet use during COVID-19 lockdown among young people in low-and middle-income countries: role of psychological wellbeing. Addict. Behave. Rep. 14, 100379 (2021) 23. Orben, A., Przybylski, A.K.: The association between adolescent well-being and digital technology use. Nat. Hum. Behav. 3(2), 173–182 (2019) 24. Li, H., Wang, J., Wang, L.: A survey on the generalized problematic Internet use in Chinese college students and its relations to stressful life events and coping style. Int. J. Mental Health Addict. 7, 333–346 (2009) 25. Lin, L.Y., et al.: Association between social media use and depression among US young adults. Depress. Anxiety 33(4), 323–331 (2016) 26. Marino, C., Gini, G., Vieno, A., Spada, M.M.: A comprehensive meta-analysis on problematic Facebook use. Comput. Hum. Behav. 83, 262–277 (2018) 27. Woods, H.C., Scott, H.: # Sleepyteens: Social media use in adolescence is associated with poor sleep quality, anxiety, depression and low self-esteem. J. Adolesc. 51, 41–49 (2016) 28. Chen, I.H., et al.: Comparing generalized and specific problematic smartphone/internet use: longitudinal relationships between smartphone application-based addiction and social media addiction and psychological distress. J. Behav. Addict. 9(2), 410–419 (2020) 29. Garrido, S., et al.: Young people’s response to six smartphone apps for anxiety and depression: focus group study. JMIR Mental Health 6(10), e14385 (2019) 30. Toscos, T., et al.: Teens using screens for help: impact of suicidal ideation, anxiety, and depression levels on youth preferences for telemental health resources. JMIR Mental Health 6(6), e13230 (2019) 31. Moreno, M.A., Jelenchick, L., Cox, E., Young, H., Christakis, D.A.: Problematic internet use among US youth: a systematic review. Arch. Pediatr. Adolesc. Med. 165(9), 797–805 (2011) 32. Kreski, N., et al.: Social media use and depressive symptoms among United States adolescents. J. Adolesc. Health 68(3), 572–579 (2021) 33. Primack, B.A., Shensa, A., Sidani, J.E., Escobar-Viera, C.G., Fine, M.J.: Temporal associations between social media use and depression. Am. J. Prevent. Med. 60(2), 179–188 (2021) 34. Dhir, A., Yossatorn, Y., Kaur, P., Chen, S.: Online social media fatigue and psychological wellbeing: a study of compulsive use, fear of missing out, fatigue, anxiety and depression. Int. J. Inform. Manag. 40, 141–152 (2018) 35. Shensa, A., Sidani, J.E., Escobar-Viera, C.G., Switzer, G.E., Primack, B.A., Choukas-Bradley, S.: Emotional support from social media and face-to-face relationships: associations with depression risk among young adults. J. Affect. Disord. 260, 38–44 (2020) 36. Bashir, H., Bhat, S.A.: Effects of social media on mental health: a review. Int. J. Indian Psychol. 4(3), 125–131 (2017)

A Review on the Impacts of Social Media on the Mental Health

195

37. Memon, A.M., Sharma, S.G., Mohite, S.S., Jain, S.: The role of online social networking on deliberate self-harm and suicidality in adolescents: a systematized review of literature. Indian J. Psych. 60(4), 384 (2018) 38. Ivie, E.J., Pettitt, A., Moses, L.J., Allen, N.B.: A meta-analysis of the association between adolescent social media use and depressive symptoms. J. Affect. Disord. 275, 165–174 (2020) 39. Watabe, M., et al.: Relationship between trusting behaviors and psychometrics associated with social network and depression among young generation: a pilot study. PLoS ONE 10(4), e0120183 (2015) 40. Karim, F., Oyewande, A.A., Abdalla, L.F., Ehsanullah, R.C., Khan, S.: Social media use and its connection to mental health: a systematic review. Cureus 12(6), 15 (2020)

Factor Influencing Online Purchase Intention Among University Students in Nepal Deepesh Ranabhat1,2(B)

, Sujita Adhikari2

, and Narinder Verma1

1 Faculty of Management Sciences, Shoolini University, Bajhol, Himachal Pradesh 173229,

India [email protected] 2 Faculty of Management Studies, Pokhara University, Pokhara-30, Kaski, Nepal

Abstract. The popularity of online shopping has grown in recent years as it provides benefits for both businesses and consumers. The purpose of this research is to determine the factors that influence online purchase intention among university students in Nepal. A total of 385 students are taken for data collection. The researchers use a structured questionnaire using Likert scale statements to collect the data. Both descriptive and inferential analysis such as frequency analysis, exploratory factor analysis, and structural equation modelling are applied. The result of factor analysis finds six factors: effort expectancy, online purchase intention, security and privacy, performance expectancy, social influence, and facilitating conditions related to online purchases. The structural equation modelling finds security and privacy, and social influence have a positive and statistically significant impact on online purchase intention. These findings emphasize the need of ensuring the security and privacy of customer data and leveraging social influence in marketing strategies to increase online purchase intention. Keywords: Online purchase intention · Security and privacy · Social influence · UTAUT model · University students · Nepal

1 Introduction Marketing is the process of addressing consumers’ needs more effectively and efficiently through better products and services, better prices, and improved access and delivery. Consumers are the focus of all marketers, who aim to understand and satisfy their needs in new and better ways. In recent years, the rise of online shopping has provided additional opportunities for marketers to reach consumers, as the internet has changed the way of shopping for goods and services to customers. Online shopping is a form of electronic commerce in which customers use web browsers to make purchases from retailers on the internet. It provides convenience and cost savings by eliminating the need to travel to physical stores and also provides more options and information for comparing products and prices. The growth of e-commerce is significant, with worldwide retail e-commerce sales projected to reach 7.4 trillion dollars by 2025 as per the report of Statista in 2021. The origins of Nepalese e-commerce can be traced back to 1999 when it was primarily used by Nepalese residing in the United States to send gifts to their friends and family © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 196–206, 2024. https://doi.org/10.1007/978-3-031-50158-6_20

Factor Influencing Online Purchase Intention Among University

197

in Nepal. At that time, internet access was limited and costly. However, the situation has changed considerably in recent years. According to the Nepal Telecommunications Authority (NTA), 91% of the country’s population now has internet connectivity, and this figure is steadily rising. According to data, most people access the internet through their mobile phones, making up 65.68% of all internet as of mid-June. There are over 12.052 million 3G internet users, with 9.590 million using Nepal Telecom and 2.462 million using Ncell. Additionally, 25.27% of users have access to fixed broadband internet, including 785,000 with Asymmetric Digital Subscriber Line (ADSL) and 556,000 with Fiber to the Home (FTTH). As internet access has become more widely available and affordable, the popularity of online shopping has grown rapidly in Nepal. Many online stores have been established in Nepal, including SastoDeal, Daraz, Muncha, SmartDoko, OkDam, Oldpinch.com, Esewapasal, Socheko.com, Thulo.com, RaraMart, Hamrobazar, Nepbay, Foodmandu, Gyapu, Meroshopping, Merokirana.com, etc. With the spread of COVID-19 and resulting lockdowns, the demand for online shopping has surged, with the number of online transactions reaching Rs. 4.93 trillion in mid-October and mid-November 2021, compared to Rs. 2 trillion in the same period last year as per the report of Central bank of Nepal. However, there are many issues related to online shopping such as the possibility of receiving a product that is different from what was advertised, limited shipping options, risk of privacy, etc. Despite some disadvantages, online shopping is seen as a convenient and cost-effective alternative to traditional retailing. It is growing in popularity among young people as they can shop from a device rather than visiting physical stores. As online purchase is increasing in Nepal, it is very important to find out the factors influencing it. So, the researchers’ main focus in this study is to find out the factors that influence online purchase intention among university students.

2 Literature Review Online shopping refers to the act of purchasing goods or services through the internet using a web browser [1]. Purchase intention is an individual’s willingness to buy a product [2, 3]. Different models have been proposed by various researchers on technology adoption including the technology acceptance model (TAM), theory of reasoned action (TRA), innovation diffusion theory (IDT), and unified theory of acceptance and use of technology (UTAUT). Among these, the UTAUT theory is considered to be a comprehensive and integrated model. It is based on eight theories and models that explain the acceptance of technology [4]. Different researchers have used UTAUT to measure the impact of user psychology on online purchase intention in various contexts. San Martín and Herrero [5] employed the UTAUT to investigate the impact of psychological factors on the intent to make online purchases for rural tourism, [6] studied online purchase intention in Vietnam, and [7] conducted a study on travellers’ intention to make a purchase of tourism products directly through online travel intermediaries (OTI) websites. These studies found four important drivers of online intention are performance expectation, effort expectation, social influence, and facilitating factors. Likewise, [8] found social influence and performance expectancy have positively influenced purchase intentions for online food delivery services. Tran and Nguyen [9]

198

D. Ranabhat et al.

found security has a significant positive impact on online shopping intention. Tandon and Kiran [10] indicates that several factors including performance expectation, effort expectation, social influence, hedonic motivation, security and privacy, and ease of ordering play a significant role in driving online shopping. This shows that several studies have used UTAUT to investigate online purchase intention in different contexts. This study also aims to understand the underlying factors that drive university students’ online purchase intention and to identify areas where improvements can be made to increase online purchase intention based on the UTAUT model. Figure 1 presents the model of the study.

Fig. 1. Model of the study

3 Research Methodology The researchers conducted a quantitative study that followed a descriptive and analytical research design. The population of interest was students of universities in Nepal and 385 respondents were taken for this study. The data was collected through a structured questionnaire using Likert scale statements to measure the independent and dependent variables. Both descriptive and inferential analysis such as frequency analysis, exploratory factor analysis (EFA) and structural equation modelling (SEM) were applied to examine the impact of independent variables on the dependent variable using SPSS Amos. The researcher used Cronbach Alpha and Composite Reliability (CR) to ensure reliability, Average Variance Extracted (AVE) for construct validity, and Fornell and Larcker criteria for discriminant validity.

4 Data Analysis and Results 4.1 Socio-demographic Profile The demographic information of the 385 survey participants, including their gender, marital status, ethnic group, field of study, level of education, monthly family income, and expenditure is presented in Table 1. The majority of respondents were female (64.7%) and unmarried (90.9%). The ethnic group was mostly composed of Brahmin (54.8%), followed by Chhetri (16.6%), Janajati

Factor Influencing Online Purchase Intention Among University

199

Table 1. Socio-demographic characteristics Variables

Categories

Frequency

Gender

Male

136 (35.3)

Female

249 (64.7)

Marital status

Married

35 (9.1)

Unmarried

350 (90.9)

Ethnic group

Brahmin

211 (54.8)

Chhetri

64 (16.6)

Janajati

68 (17.7)

Others

42 (10.9)

Education status

Bachelors

Faculty

Management

Masters and above

Monthly income of family

Monthly expenditure of family

351 (91.2) 34 (8.8) 153 (39.7)

Humanities

48 (12.5)

Health and allied sciences

56 (14.5)

Science and technology

43 (11.2)

Others

85 (22.1)

Up to Rs. 25,000

59 (15.3)

Rs. 25,001 to Rs. 50,000

145 (37.7)

Rs. 50,001 to Rs. 75,000

84 (21.8)

Above Rs. 75,000

97 (25.2)

Up to Rs. 25,000

134 (34.8)

Rs. 25,001 to Rs. 50,000

164 (42.6)

Rs. 50,001 to Rs. 75,000

46 (11.9)

Above Rs. 75,000

41 (10.6)

Total

385 (100)

(17.7%), and others (10.9%). The majority of participants were in management (39.7%), followed by humanities (12.5%), health and allied sciences (14.5%), and science and technology (11.2%). Most respondents were studying a bachelor’s degree (91.2%), with only 8.8% studying a master’s degree or higher. The majority of participants had a monthly family income between NPR 25,001 and 50,000 (37.7%) and monthly family expenditure between NPR 25,001 and 50,000 (42.6%). 4.2 Exploratory Factor Analysis (EFA) The EFA was run with 30 items related to the online shopping intention. All the communalities were found more than 0.40, which is a minimum acceptable value for large

200

D. Ranabhat et al.

sample size above 200 (Hair et al., 2019). However, five items were loaded in more than one factor, these items were removed. Finally, the EFA result was found with 25 items. Kaiser–Meyer–Olkin (KMO) and Bartlett’s tests were used to measure the appropriateness of a dataset for factor analysis. The KMO value is 0.875 which is considered as good for factor analysis, as it is more than the generally accepted value of 0.60. Likewise, Bartlett’s test of sphericity indicates that the variables are correlated with each other and are suitable for factor analysis (sig. < 0.01). The result of EFA is presented below. Table 2 lists the items that were included in the survey, along with the initial factor loading and the extracted factor loading for each item. In general, loading > 0.5 is considered as good, and loading more than 0.40 is also adequate for a sample size above 200 [11]. From the table, we can see that most of the items have a loading value of above 0.5 and ranges from 0.452 to 0.727, indicating that the items are strongly associated with the corresponding factor and are adequate for performing factor analysis. Table 3 presents the results of a factor analysis which is based on an eigenvalue of more than 1.0. A total of six factors were identified in this study, which explains 60.21% of the variance. The factors identified in this analysis are Effort Expectancy (EE) explains 12.73% of the variance, Online Purchase Intention (OPI) explains 11.798% of the variance, Security and Privacy (SP) explains 10.473% of the variance, Performance Expectancy (PE) explains 10.152% of the variance, Social Influence (SI) explains 8.235% of variance, and Facilitating Conditions (FC) explains 6.819% of the variance. Each factor is composed of several items, which are listed in the “Items” column. Additionally, the Cronbach Alpha values of more than 0.60 suggest that the items within each factor are internally consistent. 4.3 Structural Equation Modelling This includes a measurement model for validating a set of measurement items and a structural model for testing associations between the variables in the study. Measurement Model Confirmatory factor analysis (CFA) is run in measurement model. Initially, confirmatory factor analysis was run with all 25 items extracted from EFA, and model fitness was attained with 23 items. The result of measurement model is presented below. Table 4 presents fit indices for a measurement model, along with their calculated values. The indices include both absolute fit indices and incremental fit indices. The calculated CMIN/DF value is 2.146, which is below 3, indicating a well-fitted model. Similarly, the GFI is 0.907, NFI is 0.848, CFI is 0.911 and RMSEA is 0.055. Though the NFI is less than the required value of 0.90, it is near 0.90 and the calculated values of GFI and CFI are above 0.90 and RMSEA is < 0.08 signifying a well-fitted model. Table 5 presents reliability and validity statistics for a study on online purchase intention. Cronbach’s alpha measures internal consistency. A value > 0.7 is generally considered acceptable and > 0.6 is also good for a large sample size of more than 200 [11]. In this case, all of the factors have a value > 0.6, indicating that they have good internal consistency. Composite reliability (CR) measures the reliability of a scale. In this case, all of the factors have a value greater than the minimum desired value of 0.7, indicating good reliability.

Factor Influencing Online Purchase Intention Among University

201

Table 2. Communalities Item Code Items

Initial Extraction

SP1

Online shopping system provides good protection of your personal information

1.00

0.561

SP2

You feel secure in providing sensitive information for online 1.00 transactions

0.652

SP3

Internet vendors implement security measures to protect internet buyers like you

1.00

0.537

SP4

You are confident that your personal information will not be hampered when shopping online

1.00

0.567

SP5

You have trust in online shopping

1.00

0.482

EE1

Doing online shopping and web-based online transactions is 1.00 easy to you

0.547

EE2

You can easily use the shopping websites

1.00

0.668

EE3

The instruction of shopping websites is clear and easy to understand for you

1.00

0.579

EE4

Online shopping procedures are quite simple to you

1.00

0.641

EE5

You can purchase online easily with instructions

1.00

0.587

PE1

You can save the effort of visiting stores when doing online shopping

1.00

0.698

PE2

You have lots of chances to search for useful items on internet

1.00

0.675

PE3

You are able to save time in online shopping

1.00

0.611

PE4

You do not need to visit traditional shops frequently in online shopping

1.00

0.452

SI2

When online shopping is concerned, you usually do what your friends are doing

1.00

0.550

SI4

Most of your relatives and friend recommend online purchases

1.00

0.587

SI5

Your decision to purchase goods or services online is 1.00 influenced by the experiences you have had with friends and family

0.497

SI6

Important people (family/relatives/friends) help you to purchase online when you face difficulties

1.00

0.518

FC2

You have enough understanding of the online purchase

1.00

0.615

FC4

The shopping websites are linked to your various payment methods

1.00

0.645 (continued)

The average variance extracted (AVE) measures the construct validity. A value > 0.5 is generally considered acceptable. While a value > 0.4 is also considered good

202

D. Ranabhat et al. Table 2. (continued)

Item Code Items

Initial Extraction

FC5

The suppliers provides enough support and guidance for online purchase

1.00

0.528

OPI1

You prefer to purchase online

1.00

0.707

OPI2

You keep purchasing online in the future

1.00

0.743

OPI3

You are going to purchase online more frequently if possible 1.00

0.680

OPI4

You recommend online purchases to my friends and family

0.727

1.00

if the CR is > 0.7 [11]. In this case, all of the factors have a value > 0.4, indicating that they have good construct validity. In summary, the results from the reliability and validity statistics show that all the factors have good internal consistency, reliability, and construct validity. Table 6 presents results from Fornell-Lacker’s criteria, which is a way to measure the discriminant validity of the constructs. The table shows the correlation between the latent variables (SP, EE, PE, SI, FC, and OPI), and the diagonal cells represent the square root of the AVE. From the table, it is obvious that the square root of AVE of latent variables SP, EE, PE, SI, FC, and OPI is greater than the correlation value between them. So, Fornell-Lacker’s criteria are met for all the latent variables. This suggests that the constructs have good discriminant validity, and the latent variables are distinct from each other. Structural Model After fitness of measurement model, path analysis was run to examine the impact of independent variables on online shopping intention. The result of path analysis is present in Table 7. Table 7 presents the statistical results of the association between various independent variables (SP, EE, PE, SI, and FC) and a dependent variable (OPI). Based on the table, it appears that the relationships between the factors and OPI are as follows: • The relationship between Security and Privacy (SP) and OPI is statistically significant (P-value < 0.01), with an estimate of 0.275, which suggests a positive relationship between the two. • The relationship between Effort Expectancy (EE) and OPI is not statistically significant (P-value > 0.05), with an estimate of 0.008, which suggests a weak and not statistically significant association between the two. • The relationship between Performance Expectancy (PE) and OPI is not statistically significant (P-value > 0.05), with an estimate of 0.050, which suggests a weak and not statistically significant association between the two. • The relationship between Social Influence (SI) and OPI is statistically significant (P-value < 0.01), with an estimate of 0.234, which suggests a positive relationship between the two.

Factor Influencing Online Purchase Intention Among University

203

Table 3. Result of EFA Factor

Items

Loading

% of Variance

Cronbach Alpha

Effort expectancy

EE1

0.657

12.730

0.830

EE2

0.761

EE3

0.710

EE4

0.749

EE5

0.715

OPI1

0.784

11.798

0.859

OPI2

0.816

OPI3

0.775

OPI4

0.768

SP1

0.695

10.473

0.757

SP2

0.778

SP3

0.598

SP4

0.724

SP5

0.594

PE1

0.754

10.152

0.772

PE2

0.782

PE3

0.704 8.235

0.646

6.819

0.622

Online purchase intention

Security and privacy

Performance expectancy

Social influence

Facilitating conditions

PE4

0.584

SI2

0.704

SI4

0.718

SI5

0.544

SI6

0.642

FC2

0.623

FC4

0.771

FC5

0.509

• The relationship between Facilitating Conditions (FC) and OPI is not statistically significant (P-value > 0.05), with an estimate of 0.274, which suggests a weak and not statistically significant relationship between the two.

204

D. Ranabhat et al. Table 4. Measurement model

Indices

Criteria

Calculated value

Comments

CMIN/DF

0.90

0.907

Well fitted

RMSEA

< 0.08

0.055

Well fitted

Absolute fit measures

Incremental fit measures NFI

> 0.90

0.848

Moderately fitted

CFI

> 0.90

0.911

Well fitted

Table 5. Construct reliability and construct validity Factors

Cronbach alpha

CR

AVE

Security and privacy (SP)

0.69

0.75

0.43

Effort expectancy (EE)

0.83

0.84

0.52

Performance expectancy (PE)

0.68

0.73

0.48

Facilitating condition (FC)

0.62

0.72

0.46

Social influence (SI)

0.65

0.75

0.43

Online purchase intention (OPI)

0.86

0.86

0.61

Table 6. Discriminant validity- Fornell-Lacker’s criteria SP

EE

PE

SI

FC

SP

0.656

EE

0.408

0.719

PE

0.235

0.690

0.695

SI

0.355

0.365

0.374

FC

0.557

0.656

0.638

0.581

0.678

OPI

0.526

0.420

0.382

0.513

0.600

OPI

0.656 0.778

Factor Influencing Online Purchase Intention Among University

205

Table 7. Result of path analysis Relationship

Estimate

S.E

C.R

P-value

Remarks

SP → OPI

0.275

0.117

3.194

0.001

Rejected

EE → OPI

0.008

0.122

0.077

0.938

Not rejected

PE → OPI

0.050

0.162

0.45

0.653

Not rejected

SI → OPI

0.234

0.097

2.762

0.006

Rejected

FC → OPI

0.274

0.201

1.854

0.064

Not rejected

5 Conclusion This study was carried out to identify the variables influencing university students’ intentions to make online purchases. The factor analysis identified six factors: Effort Expectancy (EE), Online Purchase Intention (OPI), Security and Privacy (SP), Performance Expectancy (PE), Social Influence (SI), and Facilitating Conditions (FC) related to online purchase. Based on the result of structural equation modelling it is found that security and privacy, and social influence have a positive and statistically significant impact on online purchase intention while effort expectancy, performance expectancy, and facilitating conditions have a weak and not statistically significant relationship with online purchase intention. This finding is consistent with [8], who discovered a positive impact of social influence on online shopping intention, [9] who found positive impact of security on online shopping intention and [10] who discovered significant role of several factors including social influence and security and privacy in driving online shopping. This study concludes that online purchase intention among university students can be increased by ensuring the security and privacy of the customers’ personal and financial information. Similarly, the recommendations and experienced shared by important people like family and friends greatly influence an individual’s decision to purchase online. This study suggests that merchants should ensure the security and privacy of customer data in online shopping systems and leverage social influence in marketing strategies to increase online purchase intention among university students in Nepal.

References 1. Aldhmour, F., Sarayrah, I.: An investigation of factors influencing consumers’ intention to use online shopping: an empirical study in south of Jordan. J. Internet Bank. Commer. 21(2), 6393 (2016) 2. Tirtiroglu, E., Elbeck, M.: Qualifying purchase intentions using queueing theory. J. Appl. Quant. Methods 3(2), 167 (2008) 3. Raza, M.A., Ahad, M.A., Shafqat, M.A., Aurangzaib, M., Rizwan, M.: The determinants of purchase intention towards counterfeit mobile phones in Pakistan. J. Public Adm. Gov. 4(3), 1–19 (2014) 4. Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D.: User acceptance of information technology: toward a unified view. MIS Q. 14, 425–478 (2003)

206

D. Ranabhat et al.

5. San Martin, H., Herrero, A.: Influence of the user’s psychological factors on the online purchase intention in rural tourism: integrating innovativeness to the UTAUT framework. Tour. Manag. 33(2), 341–350 (2012). https://doi.org/10.1016/j.tourman.2011.04.003 6. Doana, T.-T.: Factors affecting online purchase intention: a study of Vietnam online customers. Manag. Sci. Lett. 10(10), 2337–2342 (2020). https://doi.org/10.5267/j.msl.2020.2.031 7. Abd Murad, S.M., Aziz, N.A.: Examining factors influencing travellers purchase intentions via online travel intermediaries websites: a conceptual model. Int. J. Econ. Res. 14(2), 289–304 (2017) 8. Hong, C., Choi, E.-K.C., Joung, H.-W.D.: Determinants of customer purchase intention toward online food delivery services: the moderating role of usage frequency. J. Hosp. Tour. Manag. 54, 76–87 (2023). https://doi.org/10.1016/j.jhtm.2022.12.005 9. Tran, V.D., Nguyen, T.D.: The impact of security, individuality, reputation, and consumer attitudes on purchase intention of online shopping: the evidence in Vietnam. Cogent Psychol. 9(1), 5530 (2022). https://doi.org/10.1080/23311908.2022.2035530 10. Tandon, U., Kiran, R.: Study on drivers of online shopping and significance of cash-ondelivery mode of payment on behavioural intention. Int. J. Electron. Bus. 14(3), 212–237 (2018). https://doi.org/10.1504/IJEB.2018.095959 11. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C.: Multivariate data analysis 87(4) (2019)

Effect of Thin Polymer Interlayers in the Spindle-Bearing Joint on the Stiffness and Durability of Spindle Bearing Assemblies of Mills A. S. Kononenko1

, T. A. Kildeev1

, Ignatkin I. Yu2(B)

, and N. A. Sergeeva2

1 Bauman Moscow State Technical University, Moscow 105005, Russia 2 Russian State Agrarian University – Moscow Timiryazev Agricultural Academy,

Moscow 127550, Russia [email protected]

Abstract. The aim of the work is to develop a scientifically justified method of calculating technological parameters of the process of creating a polymer interlayer in the spindle-bearing joint to improve the service life of the spindle bearing assembly of mills. Mathematical relationship between the stiffness of the spindle bearing assembly and physical and mechanical properties and thickness of the polymer interlayer in the spindle-bearing joint has shown that it is possible to reduce the level of deformation by increasing the elastic modulus of polymer. To investigate the effect of nanofillers on the elastic modulus of polymer, a method based on estimating the displacement of the moving part of a shaft-bearing sample under uniformly increasing load was used. A method of comparative runningin tests of spindle bearing assemblies with polymer nanocomposite interlayers between the spindle and bearing inner rings has been developed. It is shown that nanofillers positively influence the elastic modulus of anaerobic polymer and the use of nanocompositions in fixing the spindle-bearing joint increases its durability by more than 10%. Keywords: Polymer interlayer · The spindle bearing assembly · Nanocomposition · Bearing · Beam on the elastic foundation · Stiffness · Durability

1 Introduction The spindle bearing assembly is one of the most critical units of any machine tool providing the main cutting motion, which is rotation of the fixture with a workpiece or with the tool. In operation, such components of the spindle bearing assembly as bearings (75%) and spindle (9%) most frequently break out, the share of the fixing system breakdowns accounts for 6%, and that of all other elements—no more than 4% of the total number of failures [1, 2]. A promising technology for improving the durability of the spindle bearing assembly is the use of modified anaerobic polymer compositions. The essence of the method © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 207–216, 2024. https://doi.org/10.1007/978-3-031-50158-6_21

208

A. S. Kononenko et al.

consists in creating a polymer interlayer between the spindle mounting surfaces and the bearing inner rings. At the same time the modification of polymer compositions with nanosize metal oxide particles increases their elasticity modulus, accelerating the polymerization process and providing the higher stiffness of the joint in comparison with the initial composition. It is important to note that polymerization of anaerobic composition occurs immediately after assembling. For mill spindle bearings, an interference fit of 0–5 μm should be ensured when mounted on the spindle (depending on the technical requirements specified by the manufacturer). Anaerobic nanofill compounds are also suitable for the interference fit. As a result, the effect of fretting corrosion on the assembly durability is reduced and twisting of the mating parts is prevented, thus avoiding premature failure of the spindle mounting surface and the consequent costs for its restoring or producing a new spindle [3, 4]. The wear on the spindle mounting surfaces can be up to 15 μm. Depending on the stiffness requirements, polymeric nanocomposites can be used to fix a spindle bearing assembly as an independent repair technology or in combination with electromechanical machining, an additional repair part, chrome plating or surfacing. It is necessary to understand that applying of polymeric materials having a negligibly small elasticity modulus to the gap between bearing rings and spindle mounting surfaces in comparison with structural carbon steels, which are the basic material for spindle manufacturing [5, 6], will inevitably reduce the stiffness of the spindle bearing assembly and increase deformations under the influence of operating temperatures and loads, and, consequently, worsen the machining accuracy and quality. A key decision factor in selecting polymer nanocomposites and determining mounting surface restoration requirements is the predicted stiffness of the spindle-polymer-bearing joint. The optimal solution is to select repair polymer compositions with the required physical-mechanical and elastic-deformation properties [7–9].

2 Effect of the Polymer Interlayer on the Stiffness of the Spindle-Bearing Joint Let’s describe a mathematical model of interaction of a polymer interlayer with elements of the spindle bearing assembly. The spindle is mounted in the inner rings of the bearing supports with no gap. In turn, the outer rings of the bearings are rigidly fixed in the mounting bores of the stationary housing. A radial force is applied to the front end of the spindle which gives rise to a radial reaction in the front bearing support. The radial displacement of the spindle axis from the center of the housing bore is known and equals 0 . In the other case, the spindle is mounted in the inner rings of the bearing supports with a gap which is filled with solid polymer—i.e. the thickness of the polymer interlayer is equal to the gap. For the front support, its thickness should be considered thin compared to the spindle thickness and equal to h. The elasticity modulus of the polymer is considered equal to E p . The radial misalignment  of the spindle axis in relation to the center of the housing bore must be determined for the front bearing if there is a gap filled with a polymer interlayer between the spindle and the bearing inner rings.

Effect of Thin Polymer Interlayers in the Spindle-Bearing

209

In solving the above mathematical model, the spindle is supposed to be mounted in the bearings without misalignment and all the mating surfaces are considered to be absolutely smooth, and the load is transmitted through the central rolling element. It is supposed that the materials and geometric characteristics of the spindle and the bearing are known [10]. For the case where the spindle is mounted in the bearing supports with a gap filled with a polymer interlayer, we will assume that the total displacement in the bearing  is the sum (1) of the deformation of the bearing 0 and the polymer interlayer y from the applied load:  = 0 + y,

(1)

The radial displacement ratio (2) can be expressed from the formula (1) by mathematical transformations:  +y y = =1+ , 0 0 0

(2)

Let’s assume that the polymer interlayer is thick enough, i.e. the deflections that arise are little compared to the thickness of the polymer interlayer, which operates within the elasticity limits. The roughness of the mating surfaces is much lower than the thickness of the polymer interlayer. Let’s consider the plastic interlayer as an elastic foundation. Accordingly, let’s consider the inner ring of the bearing as a beam resting on the elastic foundation. The beam on the elastic foundation deforms under the action of a concentrated force. The deformation of the beam at the central point is described by formula (3): y0 =

 −1  −βl1 −2βl1 P + 2P e + cos βl e + cos 2βl βl + 2P 2βl (sin ) (sin ) 0 1 1 1 2 1 1 , 8EJ β 3 (3)

where E—elasticity modulus of the bearing material; J—moment of inertia of the beam cross-section (outer ring of the bearing); l1 —distance between the balls; P0 , P1 , P2 —load of the central, first and second ball, respectively; β—coefficient defined by formula (4):  β=

4

k , 4EJ

where k—modulus of subgrade reaction, k = k p ·b; k p —compliance ratio, N/mm3 ; b—width of the inner ring of the bearing, mm; E—elasticity modulus of bearing material, N/mm2 ;

(4)

210

A. S. Kononenko et al.

J—moment of inertia of the bearing ring cross-section, mm4 . The modulus of subgrade reaction k p is determined from Winkler formula (5) [11]: kp =

P , S

(5)

where P—pressure at the surface of the elastic foundation; S—deflection (total settlement) of the compressible elastic foundation. The deflection of polymer interlayer under pressure P can be determined from formula (6):    P 1 + μp 1 − 2μp hp   S= , (6) 1 − μp Ep where μp and E p are the averaged Poisson’s ratio and the elasticity modulus of polymer coating material within compressible thickness hp , respectively. The distortion ratio of the polymer interlayer for different thicknesses and different elasticity moduli of compositions can be represented as (7) by mathematical transformation of formulas (3), (4), (5) and (6):  

   3 4EJ 1 + μp 1 − 2μp hp P0  4    = 0 + , (7) 8EJ b 1 − μp Ep Figure 1 shows the response surface of the maximum deformation of the polymer interlayer under 5 kN load as a function of the E p and hp values.

Fig. 1. Response surfaces of maximum deformation value of polymer interlayer under 5 kN load as a function of E p and hp values.

The analysis of the presented response surfaces shows that increasing the elasticity modulus of the polymer interlayer leads to a hyperbolic decrease in its deformation at constant thickness and applied load. At the same time, the dependence of the interlayer deformation on its thickness is expressed as a law, which is close to linear. This observation suggests that increasing the elasticity modulus of polymer compounds is the most effective way of ensuring the stiffness of the spindle-polymer-bearing joint.

Effect of Thin Polymer Interlayers in the Spindle-Bearing

211

3 Increasing the Elasticity Modulus of Polymer Compositions 3.1 Materials and Methods It should be noted that modifying original polymer compositions with nanosize particles leads to a comprehensive improvement of their characteristics, including the elasticity modulus. Specific features of operation processes of the spindle bearing assembly require careful selection of polymer compositions and fillers [12, 13]. Taking into account that the gap between bearing inner ring and worn spindle mounting surface can be up to 5…15 μm, preference should be given to anaerobic sealants designed for joints with little gaps and even with slight interference. The dispersion of nanosize metal particles in the volume of the anaerobic adhesive is a catalyst for the polymerisation process. Assembling a bearing unit of a spindle head can be a challenge because of the particular design features of the spindle bearing assembly, resulting in extended service life of the finished repair compound. Given that spindle bearing assemblies are subject to periodic maintenance that involves disassembling, low-strength retainer compounds are the material of choice. Disassembling of such joints can be carried out without heating or at minimum short-term heating up to 200 °C. Loctite 601 spindle-sleeve retainers manufactured by Henkel (England) [14] and Unigerm-7 manufactured by AO NII POLYMEROV (Russia) [15] were selected for further investigation. They polymerize in the absence of air oxygen in small gaps between metal surfaces and ensure reliable fixation of joints that operate under high loads and vibrations. In the case of interaction of the polymer matrix with nanosize particles (hereinafter— NSPs), filler particles may actively participate in cross-linking processes, forming additional units or centers of structure formation, around which oriented layers of the polymer matrix with a dense packing of components comprising the nanocomposite are formed. Such assemblies can withstand extreme operating conditions without fracture [16]. Elastic properties of polymer nanocomposites can be increased due to the fact that at low volume content the NSPs move freely with the polymer matrix, which is freely stretched. Aluminum oxide (Al2 O3 ) and silicon oxide (SiO2 ) nanopowders were used as fillers in the polymer compositions. The essence of the method of assessing the elasticity modulus of compounds is to determine displacements in the boundaries of load variation in a sample under compression, the scheme of which is given in Fig. 2. The material used to manufacture the samples is steel 45. The roughness of the working surfaces is Ra = 3.2 μm. The inner and outer parts of the samples were selected so that the test compound had a thickness of 0.05 mm. After degreasing, a thin layer of sealant was applied to the working surface of the encompassed cylindrical part, and then it was placed in the second part of the sample. The elasticity modulus was assessed under normal conditions on the INTRON 600 DX hydraulic bursting machine after full polymerization of the samples, at a loading rate of 5 kN/min until the polymer layer was broken. The equipment is capable of

212

A. S. Kononenko et al.

Fig. 2. Schematic diagram of the sample: 1—pin; 2—polymer interlayer; 3—ring.

recording sample displacement from the applied loads, thus making it possible to assess the elasticity modulus of the compounds. The elasticity modulus for each sample was determined from formula (8): Ei =

Fi , S · xi /l

(8)

where Fi —load change of the i-th sample; Δx i —displacement of the i-th sample in the boundaries of the load change, mm; S—area of the polymer interlayer; l—original joint length. The results were assessed according to [17, 18]. 3.2 Results and Discussion Table 1 shows the results of the studies of the effect of nanofiller concentration on the elasticity modulus of nanocompositions based on sealants Unigerm-7 and Loctite-601. Table 1. Test results. Young’s modulus, MPa

Coctav

Composition with SiO2

Composition with Al2 O3

Concentration, %

0

0.5

1

1.5

0.5

1

1.5

Unigerm-7

171.2

181.0

203.8

238.1

246.7

289.2

290.9

254.6

214.3

281.0

266.7

369.8

300.0

280.0

214.3

340.0

272.1

246.2

314.3

307.7

261.5

217.6

248.2

257.2

323.2

249.3

274.9

257.1

216.8

242.9

274.8

240.0

273.9

275.0

200.0

223.5

271.4

242.8

200.0

342.9

280.0

314.3

Loctite-601

The analysis of experimental data showed that mixing anaerobic sealant Unigerm-7 with SiO2 nanopowder at concentration of 1.0% increased the elasticity modulus of

Effect of Thin Polymer Interlayers in the Spindle-Bearing

213

the composition by 18.2% (from 213.5 to 252.3 MPa). Adding nanopowder Al2 O3 at concentration of 0.5% raised the elasticity modulus of composition by 45.3% (from 213.5 to 310.3 MPa). Modification of a foreign compound Loctite-601 with SiO2 and Al2 O3 nanopowders at analogous concentrations enabled an increase of the elasticity modulus of the composition by 17.8 (from 219.3 to 258.2 MPa) and 31.6% (from 219.3 to 288.7 MPa) respectively. Thus, applying nanoparticles to the composition of anaerobic polymers enhances the elastic properties of the latter. The study of the extremes of the polynomial trend line equations constructed on the basis of experimental data allowed for calculation of optimum proportions of mixtures. For Loctite-601 the optimum concentrations of Al2 O3 and SiO2 nanopowders were 0.41 and 1.26%, respectively. For Unigerm-7 the optimum concentrations of Al2 O3 and SiO2 nanopowders were 0.49 and 1.38% respectively.

4 Study of the Durability of the Spindle-Polymer-Bearing Joint 4.1 Materials and Methods The development of new technologies requires validation under near operational conditions. Therefore, comparative running-in tests were carried out on spindle bearing assemblies with and without polymer nanocomposite interlayers between the spindle and bearing inner rings. When determining the condition of the bearing supports, the following indicators were taken into account: bearing noise, vibration, lubrication condition and operating temperature, the uncontrolled growth of which at the constant running speed indicates the beginning of the catastrophic wear of the assembly, i.e. the assembly has reached the limit state. The tests were performed on an experimental running-in bench, the general view of which is shown in Fig. 3.

Fig. 3. General view of the bench for comparative running-in tests of bearing units: 1—fabricated frame; 2—radial ball bearings; 3—tested assembly; 4—electric motor; 5—belt transmission; 6— eccentric.

Prefabricated aluminum profile frame 1 is mounted on the rubber isolation mounts and consists of two parts, one of which is fitted with an electric motor 4 and the other

214

A. S. Kononenko et al.

with the tested assembly. Transmission of the rotation from the electric motor shaft to the tested assembly is effected by means of belt transmission 5. Tested assembly 3 is a spindle made of steel 45, mounted on two self-centering self-aligning radial ball bearings 2. Eccentric 6 which can be fitted with up to three loads of different masses is installed at the front end of the spindle. The eccentric was selected with such parameters that the permissible residual unbalance can be exceeded by at least 100 times. The permissible residual unbalance was calculated according to formula (9) according to GOST ISO 1940-1-2007:   eper ω m , (9) Uper = 1000 · ω where U per —value of the permissible residual unbalance, g·mm; eper ω—value of balancing grade, mm/s; m—rotor mass, kg, kg; ω—angular velocity, corresponding to the maximum running speed, rad/s. Angular velocity, corresponding to a running speed of 3000 rpm, equals 314.2 rad/s. The weight of the spindle with a length of 500 mm and a diameter of 20 mm is 1.23 kg. The balancing grade is 2.5 mm/s according to GOST ISO 1940-1-2007. The permissible residual unbalance value for the tested assembly is 9.79 g·mm. It was sufficient to position a 25 g weight at an offset of 40 mm from the spindle axis to ensure the accelerated test conditions. The time was fixed with a stopwatch. The total number of cycles N was calculated as a product of the time T (min) and the running speed ω 3000 (rpm). In operation, the bearing temperature to an accuracy of ± 0.5 °C was checked with a CEM DT-8806H infra-red pyrometer, at least once every 0.5 h, and the pyrometer readings were recorded in the test report, when the temperature rose steadily for 1.5 h, the test was interrupted. Noise level was checked by means of digital noise meter MEGEON 92135 with 0.1 dB resolution and a measuring range of 30…130 dB. 4.2 Results and Discussion For the first assembly, the spindle was installed in the front support with an interference of 0.01 mm. For the second assembly, the shaft was installed in the front bearing with a 0.1 mm gap, filled with a polymer nanocomposition consisting of Loctite-601 and nanosize silicon oxide particles at a concentration of 1.3%. Abrupt changes in temperature during one hour of testing up to 6 °C as well as a 20% increase in front bearing noise from the original level were recorded for the first assembly after 6.84·106 cycles and for the second assembly after 7.56·106 cycles. Durability increased by 10%. Thus, the results confirmed the positive effect of polymer interlayers on the durability of spindle bearing assemblies. Important technological aspects when assembling the spindle-bearing joint with an interference or with a wear gap of up to 10 μm include mixing a polymer nanocomposition based on Loctite-601 or Unigerm-7 with applying of 1.3% by weight of silicon oxide nanoparticles for 10 min by means of ultrasound.

Effect of Thin Polymer Interlayers in the Spindle-Bearing

215

5 Conclusion 1. The mathematical relationship has shown that increasing the elasticity modulus of the polymer interlayer with unchanged thickness and applied load leads to a hyperbolic decrease in its deformation. 2. Modification of anaerobic polymer compositions Loktait-601 and Unigerm-7 with nanosize aluminum oxide particles allows increasing their elasticity modulus up to 80%, and nanosize silicon oxide particles up to 63%. 3. The use of polymer interlayers in the spindle-bearing joint will increase the durability of the spindle bearing assembly by more than 10%.

References 1. Shesterninov, A.V.: Designing Spindle Bearing Assemblies of Mills: Textbook, p. 96. UlGTU, Ulyanovsk (2006) 2. Push, V.E.: Designing Mills, p. 391. M: Mashinostroenie (1997) 3. Latypov, R.A., Serov, A.V., Serov, N.V., Ignatkin, I.Y.: Utilization of the Wastes of Mechanical Engineering and Metallurgy in the Process of Hardening and Restoration of Machine Parts. Part 1. Metallurgist 65(5–6), 578–585 (2021). https://doi.org/10.1007/s11015-021-01193-y 4. Mikhalchenkov, A.M., Tyureva, A.A., Kozarez, I.V., Kononenko, A.S.: The effect of the concentration of components and dispersion of particles of filler of epoxy-sand composite on hardness and its relationship with abrasive wear resistance. Polym. Sci. D 14(1), 17–20 (2021) 5. Efanov, S.A.: Provision of Parametric Reliability of Repair-Engineering Equipment by Restoration of Spindle Bearing Assemblies by Polymeric Composite Materials: Phd (Eng) Thesi, p. 131 (2015) 6. Mikhal’chenkov, A.M., Torikov, V.E., Filin, Y.I.: The influence of the concentration of components of an epoxy–sandy composite on its abrasive-wear resistance. Polym. Sci. D 11, 47–49 (2018) 7. Kononenko, A.S., Khabbatullin, R.R.: Theoretical substantiation of the conditions for the applicability of deformationless fixation by means of a polymer glue for workpieces during their mechanical processing on a milling machine with computer numerical control. Polym. Sci. D 15(4), 523–528 (2022) 8. Panfilov, Y.V., Rodionov, I.A., Ryzhikov, I.A., Baburin, A.S., Moskalev, D.O., Lotkov, E.S.: Ultrathin film deposition for nanoelectronic device manufacturing/vacuum science and equipment. In: Proceedings of the 26 Conference with International Participants IOP Conference Series: Materials Science and Engineering, p. 781. IOP Publishing, New York (2020) 9. Kononenko, A.S., Ignatkin, I.Y., Drozdov, A.V.: Recovering a reducing-gear shaft neck by reinforced-bush adhesion. Polym. Sci. D 15(2), 137–142 (2022) 10. Schaeffler Group Industrial. https://www.schaeffler.com/content.schaeffler.com/en/divisions/ industrial/industrial.jsp. Accessed 16 Dec 2022 11. Gorb, A.M.: Improvement of Analytical Methods of Calculation of Industrial Floor Structures Made of Cement Concrete on the Elastic Foundation in Case of Using Local Resilience Model [Text]: PhD (Eng) Thesis, p. 140 (2009) 12. Kononenko, A.S., Ignatkin, I., Drozdov, A.V.: Restoring the neck of a reducing-gear shaft by gluing a reinforced bush. Polym. Sci. D 15(4), 529–534 (2022)

216

A. S. Kononenko et al.

13. Li, R.I., Psarev, D.N., Kiba, M.R.: Promising Nanocomposite Based on Elastomer F-40 for Repairing Base Members of Machines. Polym. Sci. D Glues Seal. Mater. 12(2), 128–132 (2019) 14. Loctite company. https://loctayt.pyc/. Accessed 28 Dec 2022 15. AO NII POLYMEROV. http://www.nicp.ru/. Accessed 28 Dec 2022 16. Mikhalchenkov, A.M., Kravchenko, I.N., Filin, Y., Kozarez, I.V., Velichko, S.A., Erofeev, M.N.: Abrasive wear mechanism of polymer composites with a dispersed filler. Refract. Ind. Ceram. 63, 174–177 (2022) 17. Tikhomirov, D., Kuzmichev, A., Rastimeshin, S., et al.: Energy-efficient pasteurizer of liquid products using IR and UV radiation. Adv. Intell. Syst. Comput. 866, 178–186 (2019). https:// doi.org/10.1007/978-3-030-00979-3_18 18. Dorokhov, A., Kirsanov, V., Pavkin, D., et al.: Recognition of cow teats using the 3D-ToF camera when milking in the “herringbone” milking parlor. Adv. Intell. Syst. Comput. 1072, 128–137 (2020). https://doi.org/10.1007/978-3-030-33585-4_13

The Use of a Nutrient Solution Containing Chelated Forms of Various Trace Elements K. Pishchaeva1(B)

, S. Muradyan1 , E. Nikulina2 and A. Saproshina1

, S. Buleeva1

,

1 UNESCO Chair in Green Chemistry for Sustainable Development, Dmitry Mendeleev University of Chemical Technology of Russia, Miusskaya sq. 9, 125047 Moscow, Russia [email protected], {pishchaeva.k.v,muradian.s.a, buleeva.s.v,saproshina.a.a}@muctr.ru 2 National Research Center “Kurchatov Institute”, sq. Academician Kurchatova, 1, 123182 Moscow, Russia

Abstract. This study tested for the first time the composition of a nutrient solution for hydroponic cultivation of daikon with chelated forms of 4 essential trace elements (Fe, Zn, Cu, Mn) with carboxyl-containing ligands—ethylenediaminetetraacetic acid (EDTA) and diethylenetriaminepentaacetic acid (DTPA). The results of the study were compared with the common GHE nutrient solution, which is widely used in hydroponics. The use of the proposed composition led to the greatest growth and weight of plants. The value of the mass of plants grown in the experimental nutrient solution was 30% more than with the GHE nutrient solution (General Hydroponics Europe, France), as well as the leaf area by 16%. In all samples, the plants did not show chlorosis, there was no yellowing of the leaves, and good turgor was observed. Thus, a micronutrient nutrient solution with carboxyl-containing ligands (DTPA, EDTA) is promising for growing daikon in hydroponic conditions and can be studied to optimize the cultivation of other agricultural products. Keywords: Hydroponic cultivation · Chelates · Daikon

1 Introduction Global food security is one of the biggest challenges facing global agriculture [1]. According to demographers, by 2050 the world’s population will be 9.5 billion people, of which more than 65% will live in urban areas [2]. To meet the projected increase in the number of people, a significant increase in crop yields is required. Given the global trend in the reduction of arable land [3] and the complexity of modern natural resource management, there is an urgent need to develop and further improve sustainable cultivation systems. Growing food using hydroponic methods is one such system. The benefits of hydroponically grown foods are numerous. These include efficient water use, limited use of pesticides, higher yields, and year-round food production [4]. Hydroponic systems provide the opportunity to grow agricultural products in non-arable regions of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 217–222, 2024. https://doi.org/10.1007/978-3-031-50158-6_22

218

K. Pishchaeva et al.

the world: arid regions or in urban areas, in the regions of the north and the Arctic Circle [5]. This line of research is relevant, since agriculture in the northern boreal [6] and Arctic regions of Russia is currently underdeveloped due to harsh climatic conditions, and the use of traditional agricultural technologies is not enough to meet the food needs of the local population. Traditional methods of infrastructure development and adaptation of cropping systems are losing their relevance, as they can affect the transformation of natural lands and carry risks for the environment of the Arctic zone. The widespread introduction of soilless growing systems (vertical farms) [7] of agricultural products is an important element in providing the population with food. Thus, studies on the creation of optimal nutrient compositions for growing test crops in hydroponic conditions are of undoubted relevance. Daikon has numerous beneficial properties and is therefore very promising for growing in hydroponics. The fruits contain many vitamins, enzymes and trace elements necessary for human health. In practice, daikon is often included in recipes for the prevention and treatment of many diseases, general strengthening of immunity [8]. For the vast majority of crops, including hydroponic daikon, there is no ideal growing solution that allows plants to reach their maximum genetic and physiological potential. New comprehensive studies are needed to study the exact requirements of plants for the implementation of a complete metabolism. Low yields, lean plants, high water and reagent costs, which are symptomatic of unsuccessful cultivation, are most often directly related to incorrect formulations and improper management of the nutrient solution [9]. At the same time, issues of sustainable agricultural intensification apply equally to hydroponic growing systems. Therefore, the issues of establishing the optimal amounts of essential nutrients for specific crops, the search, and the study of formulas that can additionally activate the internal reserves of plants continue to be relevant. This study is devoted to approbation of the optimal composition for growing daikon in hydroponic conditions using chelate forms of key trace elements (Fe, Zn, Cu and Mn) based on carboxyl-containing ligands, analysis of the effect of the composition on the growth and development of daikon.

2 Materials and Methods To study the properties of the developed nutrient solution, the daikon Baseball F1, Gavrish company, was chosen as a test culture. The hydroponic system consists of two plastic containers with a volume of 5 L filled with a nutrient solution. The first container contained GHE nutrient solution (France), the second container contained a developed analogue nutrient solution with trace elements, Cu, Zn, Mn. Such trace elements Cu, Zn, Mn were added in a chelated form based on EDTA, and Fe on the basis of DTPA. The nutrient solution contained the following concentrations of macro- and micro-elements (%): (NO3 )− —4.04; (NH4 )+ —1.71; K2 O—3.98; PO4 —6.13; CaO—1.38; MgO—0.87; (SO4 )2− —0.93; Fe—0.025; B—1.5; Cu—0.003; Zn—0.004; Mn—0.01; Mo—0.18. It has been proposed to add a complex of Fe with diethylenetriaminepentaacetic acid due to the instability of the complex of iron with EDTA in alkaline solutions. Samples of nutrient solutions were prepared at the UNESCO Chair in Green Chemistry for Sustainable Development, Dmitry Mendeleev University of Chemical Technology of Russia.

The Use of a Nutrient Solution Containing Chelated Forms

219

Daikon seeds in the amount of 120 pieces in each container were placed on a gauze napkin, which was irrigated with a nutrient solution. The containers were then placed in a climate chamber with optimal humidity and temperature control, equipped with LED lighting (Fig. 1).

Fig. 1. Daikon seeds in a growing container.

The nutrient solution was constantly aerated to maintain an oxygen concentration above 15 mg/L. The pH values were up to 5.5–6.5. The air temperature in the chamber was 24–26 °C, the humidity was maintained at 60%. The decrease in the water level in the containers due to evaporation was compensated by adding water to the maximum level. All experiments were prepared using triplicate. Sampling for measuring biometric parameters was carried out on the 30th day after germination. To determine the main growth parameters, 20 plants were selected from each container. To determine the dry weight, the samples were crushed and dried in an oven at a temperature of 60–70 °C for 3 h to a constant weight. Mathematical processing of the obtained data was carried out using the methods of variation statistics in the form of calculations of the arithmetic mean, standard deviation, coefficient of variation, and Fisher’s test using Microsoft Excel.

3 Results and Discussion During testing, changes in biometric and morphometric parameters of daikon growth and development were recorded on the 30th day (Table 1). The results showed that the selected optimal concentrations and ratios of macroand micro-elements, as well as the use of Cu, Zn, Mn based on EDTA and Fe based on DTPA in the form of chelates, made it possible to achieve a significant positive effect—an increase in the growth and biomass of daikon compared to the GHE solution.

220

K. Pishchaeva et al. Table 1. Parameters of the grown daikon

Nutrient solution

Plant height (cm)

Fresh leaf mass (g)

Dry weight of plants (g)

Dry weight of roots (g)

Leaf surface area of plants (cm2 )

GHE

14.2 ± 1.1

14.9 ± 0.015

0.67 ± 0.012

0.06 ± 0.01

1.1 ± 0.05

Experimental solution

14.7 ± 1.5

21.2 ± 0.02

0.96 ± 0.018

0.17 ± 0.008

1.3 ± 0.08

The mass of plants in the container with the experimental nutrient solution exceeded the mass of plants with the GHE solution by 30%. The leaf surface area of daikon in the experimental solution is 16% larger compared to the GHE nutrient solution. Plant height was almost the same in containers with GHE solution and developed nutrient solution. In general, the plants in all containers had a shiny leaf plate without signs of chlorosis, yellowness (Fig. 2). However, it is worth noting that the daikon grown on the experimental nutrient solution had a large developed root system (root mass was 65% higher compared to plants grown on the GHE nutrient solution).

Fig. 2. Growing daikon in containers containing nutrient solutions.

The replacement of trace elements with chelated forms with a phosphorouscontaining ligand led to a positive side technological effect—the absence of an increase in bacterial films and mucus on surfaces in contact with the solution, which is an undoubted advantage in exploitative terms. Together with an increase in the resistance of cultivated crops to water and nutrient stress, the absence or very slow growth of bacterial films significantly expand the possibilities for optimizing and intensifying hydroponic crop production.

The Use of a Nutrient Solution Containing Chelated Forms

221

4 Conclusion Thus, experimental tests have shown that a nutrient solution containing trace elements (Fe, Zn, Cu, Mn) with carboxyl-containing ligands (EDTA, DTPA) has significant potential for practical use in hydroponic systems. The mass of daikon grown in the experimental nutrient solution increased by 30% more compared to the GHE nutrient solution, leaf area by 16%. The developed nutrient solution contains micronutrients in digestible forms, which is important for the growth of the daikon. Nutrient deficiency reduces photosynthesis, causes growth inhibition, affects leaf area, and accelerates aging [11, 12]. Growing agricultural [13] products using hydroponics can contribute to the active development of the infrastructure of the Arctic regions of Russia and their development. Compared to the occurrence of growing cases, hydroponic methods cover different places and territories. The introduction of industrial systems of hydroponic cultivation in the Arctic zone will quickly provide the population with high-quality, fresh and inexpensive food all year round. At the same time, the developed optimal compositions of nutrient solutions are important for the economic efficiency of production. At the moment, research is expanding and experiments are being carried out with the cultivation of basil. Based on the proposed approach using chelate forms of these trace elements, it is planned to develop optimal nutritional compositions for growing tomatoes, strawberries, various types of greens and other economically important crops in vertical farms. Acknowledgments. The reported study was funded by MUCTR within the framework of the internal initiative grant No VIG-2022-037.

References 1. Rosegrant, M.W., Cline, S.A.: Global food security: challenges and policies. Science 302, 1917–1919 (2003) 2. Department of Economic and Social Affairs. https://www.un.org/sw/desa/68-world-popula tion-projected-live-urban-areas-2050-says-un 3. Olsson, L., et al.: Land degradation. In: Shukla, P.R., et al. (eds.) Climate Change and Land: An IPCC Special Report on Climate Change, Desertification, Land Degradation, Sustainable Land Management, Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems. IPCC, Geneva (2019) 4. Radhakrishnan, G., Upadhyay, T.K., Singh, P., Sharma, S.K.: Impact of hydroponics: present and future perspective for farmer’s welfare. Suresh Gyan Vihar Univ. Int. J. Environ. Sci. Technol. 5(2), 19–26 (2019) 5. Interreg Europe. https://projects2014-2020.interregeurope.eu/cityzen/news/news-article/ 11981/hydroponics-and-its-role-in-urban-agriculture/ 6. Altdorff, D., et al.: Agriculture in boreal and Arctic regionsrequires an integrated global approach for research and policy. Agron. Sustain. Dev. 41, 23 (2021) 7. Van Gerrewey, T., Boon, N., Geelen, D.: Vertical farming: the only way is up? Agronomy 12(1), 2 (2022) 8. Sela Saldinger, S., Rodov, V., Kenigsbuch, D., Bar-Tal, A.: Hydroponic agriculture and microbial safety of vegetables: promises, challenges, and solutions. Horticulturae 9, 51 (2023)

222

K. Pishchaeva et al.

9. Velazquez-Gonzalez, R.S., Garcia-Garcia, A.L., Ventura-Zapata, E., Barceinas-Sanchez, J.D.O., Sosa-Savedra, J.C.: A review on hydroponics and the technologies associated for medium- and small-scale operations. Agriculture 12, 646 (2022) 10. Sakamoto, M., Komatsu, Y., Suzuki, T.: Nutrient deficiency affects the growth and nitrate concentration of hydroponic radish. Horticulturae 7, 525 (2021) 11. Mu, X., Chen, Q., Chen, F., Yuan, L., Mi, G.: Within-leaf nitrogen allocation in adaptation to low nitrogen supply in maize during grain-filling stage. Front. Plant 7, 699 (2016) 12. Mu, X., Chen, Y.: The physiological response of photosynthesis to nitrogen deficiency. Plant Physiol. Biochem. 158, 76–82 (2021) 13. Hossain, K.A., et al.: Paddy: disease prediction using convolutional neural network. In: Vasant, P., Zelinka, I., Weber, G.W. (eds.) Intelligent Computing and Optimization. ICO 2021. Lecture Notes in Networks and Systems, vol. 371. Springer, Cham (2022). https://doi.org/10.1007/ 978-3-030-93247-3_27

Designing an Inventory Control System in Food and Beverage Industry Tiovitus Flomando Tunga1 and Tanti Octavia1,2(B) 1 Industrial Engineering Department, Petra Christian University, Surabaya, Indonesia

[email protected] 2 Engineer Profession Education Department, Petra Christian University, Surabaya, Indonesia

Abstract. X Corp. is a company engaged in the food and beverage industry. With the growth of the F&B industry, companies need to implement good inventory management as part of their efforts to win the supply chain competition. Good inventory management is achieved when the inventory costs incurred are at a minimum. A model or ordering system policy comparison needs to be done in an effort to achieve optimal inventory management. The continuous review method can minimize inventory costs compared to the periodic review method and the current inventory control at the company with the lowest holding and ordering costs. In 2021, this method can save up to IDR. 100,000,000 for the RM-SAC0042 type. In carrying out warehousing activities in new warehouses, companies need to implement computer systems to increase the performance of warehousing activities. A database system for stock management in the warehouse will accommodate the company in carrying out warehouse activities and make them more effective. The most optimal inventory management method, which is the continuous review method, is then implemented in the raw material warehouse database system of X Corp. Keywords: Supply chain · Inventory management · Warehouse database · Food and beverage

1 Introduction One of the most difficult issues in business, particularly in the food and beverage sector, is maintaining inventory of finished items and raw materials. X Corp., a company in the food and beverage (F&B) industry, uses a make-to-order strategy to meet customer demand. Currently, the company is having trouble keeping up with customer demand while managing its raw material stocks. Galea-Pace [1] stated that all parts of the supply chain converge in inventory, which is regarded as a basic component of supply chain management or supply chain. Many research papers have been published in the inventory management area. Kandananond [2] examined various forecasting techniques for inventory requests and assessed each one’s effectiveness using actual data. Stock et al. stated traditional logistics decisions have received a lot of attention from researchers in the field of logistics, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 223–235, 2024. https://doi.org/10.1007/978-3-031-50158-6_23

224

T. F. Tunga and T. Octavia

but more lately, they have turned their attention to collaborative models for inventory management [3]. While Wang [4] created a review of the several inventory management strategies based on supply chain management (SCM), and organized your papers using the one-, two-, and multi-echelon strategies. To establish the appropriate cycle duration and the percentage of no shortages, Wu et al. [5] consider the two important and vital factors described above in order to maximize the overall profit. Ben-Ammar et al. [6] strive to identify the appropriate order release dates for components at the final level of the BOM in order to decrease the sum of the average backlog and inventory holding costs for the finished product as well as the average inventory holding costs for components. Senthilnathan [7] implement branch and bound in the four-layer production inventory model. In order to help firms, maximize their inventory management efforts, this study suggests using ABC analysis to categorize inventory items based on their relative value. Inventory control or inventory management of raw materials is very much needed by the company to handle problems like this. Inventory is considered a core component of supply chain management and is a place where all areas of the supply chain come together [1]. In an area of intense industrial competition, inventory managers are encouraged to reduce inventory costs, improve inventory flow in the supply chain, and meet customer demands on time. Therefore, the implementation of effective inventory control is needed by X Corp to continue to meet customer demands with minimal inventory costs. In inventory control, the company also faces other challenges, such as the lack of a database that can support the implementation of inventory management. Inventory control performance can also be boosted by a database system that can speed up the flow of data information in the warehouse, such as stock, data for taking goods, and placing goods, and reduce information errors. In this study, analysis and proposals will be made regarding the design of inventory control for X Corp.

2 Methods Before creating the warehouse database, this research compared three different inventory control strategies. These techniques include continuous review, periodic review, and the company’s current approach to inventory control in 2021. 2.1 Continuous Review Step 1: Economic Order Quantity How many products and when to make a purchase are two issues that frequently arise while acquiring policies or purchasing goods. Finding the ideal order quantity with the lowest ordering expenses and storage costs is the topic of the modeling approach known as Economic Order Quantity [8]. The following is the formula for calculating the EOQ value:  2DS (1) Q∗ = H

Designing an Inventory Control System in Food and Beverage

225

where: Q* = Economic Order Quantity S = Ordering cost H = holding cost D = Need for quantity of goods Step 2: Safety stock A safety stock is an inventory that is kept to guard against fluctuating consumer demand. The corporation maintains safety stock to meet with demand that is unexpected and under/over forecasts [9]. This may depend on the wait period’s duration, desired service quality, frequency of reorders, and fluctuation of demand during the lead time. With the formula below, safety stock may be calculated. SS = Z × Sd

(2)

where: SS = safety stock Z = Service factor Sd = Standard deviation during lead time Step 3: Reorder Point Reorder or reorder point (ROP) is established based on a number of factors, including average need, material utilization, order grace period, and lead time [10]. The reorder point must also consider additional criteria, one of which is safety stock. Calculations utilizing the following formula can be used to determine whether a reorder is necessary or to determine a reorder point. ROP = (d × LT ) + SS

(3)

where: ROP = Reorder point D = Average demand per day LT = Lead time SS = Safety stock. 2.2 Periodic Review Companies can discover inventory movements by using periodic review policies or procedures. In this instance, the corporation can determine slow-moving commodities by examining the amount of inventory it has at each predetermined time interval. So,

226

T. F. Tunga and T. Octavia

businesses can lower their inventory levels [11]. These are the steps in the periodic review approach calculation: Step 1: Calculating Safety Stock The calculation of the periodic review begins with calculating the safety stock and then proceeds to the calculation of the maximum inventory. Safety stock in the periodic review method can be calculated using the following formula: SS = Z × SdR + LT

(4)

where: SS = safety stock Z = Service factor Sd R + LT = Standard deviation during lead time plus time review Step 2: Calculating Maximum Inventory. The safety stock value obtained is then used in calculating the maximum inventory or maximum inventory. Maximum inventory is the maximum amount or quantity of inventory the inventory level is below the maximum inventory, then an order must be placed to meet the maximum inventory. Maximum inventory calculation can be done using the following formula. T = d × (R + LT ) + SS

(5)

where: T = Maximum inventory D = Average demand per day R = Time review LT = Lead time SS = Safety stock.

3 Research Methodology Problem identification is the initial step in this research project. The company’s inventory control issues are discussed verbally with X Corp. to identify the concerns. Based on the interview, it was determined that the organization has issues managing raw material inventories because of changes and ambiguity in client demand. Currently, the business is preparing inventory control for three new warehouses. The types of raw materials investigated include those from the goods with the highest degree of demand, namely Tapioca Pearl products, the primary raw materials of which are RM-SAC-0042, RMCGC-0013, RM-GRI200-0011, and RM-TRB-001. The data is obtained by requesting a monthly recap of raw material usage data for 2019 to 2021.

Designing an Inventory Control System in Food and Beverage

227

The raw material usage data obtained from the company will be calculated using 3 methods: continuous review, periodic review, and current inventory control of the company. Data processing is done with Microsoft Excel software. The results of the analysis and comparison will be implemented into the inventory management database to be designed. Inventory management system design is proposed after data processing, in which two methods of raw material inventory control are carried out and analyzed. The raw material inventory control method that produces the lowest cost will be simulated in the inventory management database. The inventory management system is designed with Visual Basic for Applications in Microsoft Excel software. The design of the inventory management system is carried out in the raw material warehouse, finished goods warehouse, and oatmeal raw material warehouse. After the inventory management database design is complete, the analysis and conclusions are drawn regarding the overall research and become suggestions for future research for the company.

4 Results and Discussions 4.1 Ordering and Holding Cost Calculation One of the main variables in calculating the inventory control method is ordering cost (order cost) and holding cost (storage cost). The ordering cost is the cost incurred by the company to place an order for raw materials. The average duration of ordering raw materials by telephone is 5 min, which costs IDR 6360. The company also assigned one employee to make purchase orders for about 30 min. Thus, the total time required to place an order is 35 min. The following is the calculation of the cost of ordering raw materials. Calling cost = IDR 636 per 30 s = IDR 6360 for 5 min Salary of 1 employee = IDR4, 262, 015/month = IDR3405 per 35 min Ordering Cost = IDR 6360 + 3405 = IDR 9766 There are several regulations from suppliers regarding the minimum purchase of raw materials. In the RM-SAC-0042 and RM-TRB-001 raw materials, the minimum purchase of raw materials is a multiple of 20 tons, or 20,000. For the raw materials RM-CGC-0013 and RM-GRI200-0011, ordering raw materials can only be done in multiples of 25 kg. Holding costs are costs incurred by the company to store goods. The calculation of the company’s storage costs is carried out by considering factors such as electricity costs incurred per month, the total cost of all warehouse employees’ salaries, and warehouse capacity for all raw materials. The company spent IDR 1,000,000 for electricity costs per 20 days. Thus, the company issued IDR 1,500,000 for 1 month’s electricity costs. The company has 9 warehouse employees with a UMR Tangerang salary of IDR 4,262,015. The amount of funds issued by the company per month for the salaries of nine warehouse employees is IDR 38,358.135. The raw material warehouse can accommodate a total of

228

T. F. Tunga and T. Octavia

about 210 tons. The following is the calculation of the company’s warehouse storage costs. Holding Cost =

1, 500, 000 + 38, 358, 135 = IDR190 per kg per month 210, 000

Thus, the company’s holding cost is IDR 190 per kg per month and a one-time order fee of IDR 9766 for 5 min of call duration. The results of this cost calculation will be used in calculating inventory control methods such as continuous review, periodic review, and current inventory control. 4.2 Continuous Review Calculation These are detail calculations of continuous review method for RM-SAC-0042: 1. Economic Order Quantity The first stage in the analysis of inventory control using the continuous review method is to calculate the economic order quantity. The following is the calculation of the economic order quantity for the raw material RM-SAC-0042. Calling cost = IDR 9.766 Holding cost = IDR 190 per kg per month  ∗

Q =

2 × 11, 150 × 9766 = 1.072 kg 190

The average demand or use of RM-SAC-0042 raw materials per month from January to December 2021 is 11,150 kg. Through the calculation of the economic order quantity formula, the ideal number of orders for the RM-SAC-0042 item for one order is 1072 kg, but regarding the regulation from the supplier, the minimum order quantity that can be placed is 20,000 kg. 2. Safety Stock The calculation of safety stock is the second calculation stage before proceeding to the calculation of the reorder point or total inventory cost. The following is the calculation of safety stock. Usage variance per month = 32.424687 Variance lead time(30 days) = 38, 909, 624 Standard deviation(lead time) = 5.694 Service level = 90% Z(service factor) = 1.28 SS = Z × Sd = 1.28 × 5.694 = 7297 kg

Designing an Inventory Control System in Food and Beverage

229

The company has a 30-day lead time from ordering raw materials until the order arrives from the supplier. The service level value is obtained from information about the percentage of the company’s probability of fulfilling all customer orders. Through the calculation of safety stock, the minimum stock or inventory that must be owned by the company is 7994 kg. 3. Reorder Points The reorder point is calculated based on the safety stock value obtained and the lead time required by the company to receive inventory from the customer. The following is the calculation of the reorder point. Average usage = 446(in days) Lead time = 30 days ROP = (d × LT) + SS ROP = (446 × 30 days) + 7994 kg = 21.374 kg Through calculations, the reorder point value for the RM-SAC-0042 item was obtained at 21.374 kg. 4.3 Periodic Review Calculation These are detail calculations of periodic review method for RM-SAC-0042. 1. Safety Stock In the periodic review method, the calculation of safety stock is the first calculation stage before proceeding to the calculation of maximum inventory or total inventory cost. The following is the calculation of the safety stock with the periodic review method for the raw material item RM-SAC-0042. Time Review(R) = 15 days Lead Time(LT) = 30 days (R + LT) = 45 days Variance(45 day) = 58, 364, 436 St. dev(45 days) = 7640 Service level = 90% Z(service factor) = 1.28 Safety Stock = 1.28 × 7.640 = 9790 kg The time review shows the time interval in which the company must reorder, which is every 15 days. The service level value is obtained from information about the percentage of the company’s probability of fulfilling all customer orders. Through the calculation of safety stock, the minimum stock, or inventory item RM-SAC-0042, that must be owned by the company for the periodic review method is 9790 kg.

230

T. F. Tunga and T. Octavia

2. Maximum Inventory The results of the safety stock calculation are carried over to the maximum inventory calculation. The following is the maximum inventory calculation for RM-SAC-0042 raw materials. Average usage = 446 kg(in days) T = d × (R + LT) + SS = 446 kg × (15 days + 30 days) + 9790 kg = 2, 986, 038 kg Through calculations, the maximum inventory value for the item RM-SAC-0042 is 29,860 kg. The number of orders that must be made by the company follows the available stock at that time, namely Maximum Inventory (T)—Stock on Hand. 4.4 Total Inventory Cost Comparison The results of the calculations of the two previous methods are then simulated into raw material usage data in 2021. This aims to calculate the total inventory costs from the implementation of each method (including the current inventory control) and compare which ones are the most effective and economical.

Fig. 1. Total inventory cost comparison between each method

Based on Fig. 1, it can be seen that the continuous review method produces the minimum inventory costs for most of the raw materials, including RM-SAC-0042, RMCGC-0013, and RM-GRI200-0011. Periodic review produces higher inventory levels than continuous review. This could be due to the raw materials ordered at the end of the year under the continuous review method arriving in 2022, so there will be a significant difference in inventory levels in 2021. For raw materials RM-SAC-0042, inventory control using the continuous method can save up to IDR 110,000,000 of the inventory control carried out by the company in 2021. Orders were only made three times, while in the company’s inventory control in 2021, orders were made up to five times. In this case, the cost of the message becomes more efficient at around IDR 20,000.

Designing an Inventory Control System in Food and Beverage

231

There are substantial savings in storage costs. The company stores 39,432,807.46 kg of raw materials, while with the continuous review method, the company only needs to store 22,017,387.03 kg. This difference of 17,415,420.43 kg can save storage costs of IDR 110,297,663. In the application of the continuous review method, orders for raw materials for RM-SAC-0042 are made up of 20 tons, or 20,000 kg, and each inventory figure is 21.374 kg or less. Reviews and periodic reviews, there is a backorder cost, namely the inability to meet the needs of the use of raw materials. This happens because the use of raw materials continues, but orders for raw materials have not been received from the supplier.

Fig. 2. Main menu interface

Although the backorder cost for the continuous review method is higher, the total inventory storage with the periodic review method is much larger, so the total holding cost and the total inventory cost for the continuous review method are still the least expensive. This can also be an evaluation for the company, seeing that the total costs incurred with inventory control carried out in 2021 are the highest for the raw materials RM-SAC-0042, RM-CGC-0013, and RM-GRI200-0011. Inventory control using the continuous review method is more efficient on ordering costs and storage costs than the company’s current inventory control, with fewer orders and much less storage than the company will do in 2021. The continuous review method on raw materials RM-CGC0013 and RM-GRI200-0011 can save inventory costs up to IDR 180,000,000 of the total inventory costs arising from inventory control carried out by the company during 2021. Inventory control carried out by the company in 2021 can be considered good for RM-TRB-001 raw materials because it incurs the smallest inventory costs compared to the two simulated methods, which is IDR 33,164,505. The periodic review method of inventory control is not suitable for RM-TRB-001 raw materials because it causes a lot of storage even though the number of orders is small. The inventory control method that produces the lowest cost for each type of raw material is then implemented into the proposed database design for the raw material warehouse at the Teluk Naga plant.

232

T. F. Tunga and T. Octavia

A database is designed to accommodate warehouse activities such as item picking, putting away, stock updating, and item tracing. There are 4 important menus in the database, named main menu, item picking, item receiving, and item stock. 1. Main Menu The main menu is the first menu that will appear when the user opens the database. This menu connects with three other important menus that were mentioned before. There is also a visual layout to give the user more visual information about racks and the items stored in those racks. The interface is shown in Fig. 2. 2. Item Picking Item-picking menus are designed to accommodate item-picking activities in warehouses. The item-picking interface is shown in Fig. 3.

Fig. 3. Item picking menu interface

All data input in this menu will be exported as a picking list. A picking list is an order document for picking up goods that contains data on the slot code and the quantity that has to be taken. Based on discussions with the head of the warehouse department, it was decided that the warehouse staff did not need to know the type of item taken. The picking list is shown in Fig. 4. 3. Item Receiving The item receiving menu is intended to support item storing or putting activities. The “receiving” term is better known by warehouse staff and admin. The item receiving interface is shown in Fig. 5. 4. Item Stock

Designing an Inventory Control System in Food and Beverage

233

Fig. 4. Picking list

Fig. 5. Item receiving interface

Item stock is a menu designed to see the availability and quantity of items in the whole warehouse. This menu is also the implementation of the continuous review method. Every item that has reached the reorder level will display an alert on the menu. The item stock interface is shown in Fig. 6.

234

T. F. Tunga and T. Octavia

Fig. 6. Item stock interface

5 Conclusion X Corp. is a company engaged in the field of food and beverage. In a fiercely competitive industry, companies need to implement effective inventory management or control to ensure supply chain management performance remains optimal and can meet customer satisfaction. Good inventory control is achieved when the total inventory costs incurred are low. Based on the calculation of the reorder point model, periodic reviews, and current inventory control, each type of raw material does not cause costs that are always low in one particular method. For raw materials RM-SAC-0042, RM-CGC-0013, and RM-GRI200-0011, the method that produces the least inventory costs is continuous review. In the RM-TRB-001 raw material, the method that produces the least cost is the current inventory control of raw materials. In carrying out warehousing activities at the new warehouse of the Teluk Naga plant, companies need to implement computerization or digitization to improve the performance of warehousing activities. The design of a database system for stock management in the warehouse will accommodate companies in carrying out warehouse activities such as item picking, putting-away, and inventory recording so that they become more effective, precise, and well-organized. The results of the calculation of the most optimal raw material inventory control are then implemented in the raw material warehouse database of X Corps Teluk Naga plant. Thus, the company has a reference regarding when to place an order, what quantity of goods to order, and which shelf locations the goods can be placed on.

Designing an Inventory Control System in Food and Beverage

235

References Galea-Pace, S.: Why is inventory management in the supply chain important? (2020). https:// supplychaindigital.com/digital-supply-chain/why-inventory-management-supply-chain-imp ortant. Accessed 22 Dec 2022 Kandananond, K.: A comparison of various forecasting methods for autocorrelated time series. Int. J. Eng. Bus. Manag. 6, 18–23 (2012). https://doi.org/10.5772/51088 Williams, B.D., Tokar, T.: A review of inventory management research in major logistics journals. Int. J. Logist. Manag. 19(2), 212–232 (2008). https://doi.org/10.1108/09574090810895960 Wang, K.: The research of inventory management modes based on supply chain management. In: International Asia Conference on Industrial Engineering and Management Innovation Proceedings, pp. 1319–1329 (2013). https://doi.org/10.1007/978-3-642-38445-5_137 Wu, J., Teng, J.-T., Chan, Y.L.: Inventory policies for perishable products with expiration dates and advance-cash-credit payment schemes. Int. J. Syst. Sci. Oper. Logist. 5(4), 310–326 (2017). https://doi.org/10.1080/23302674.2017.1308038 Ben-Ammar, O., Dolgui, A., Hnaien, F., Louly, M.: Supply planning and inventory control under lead time uncertainty: a review. IFAC Proc. 46(9), 359–370 (2013). https://doi.org/10.3182/ 20130619-3-RU-3018.00592 Senthilnathan, S.: Economic order quantity (EOQ). SSRN Electr. J. (2019). https://doi.org/10. 2139/ssrn.3475239 Panja, S., Mondal, S.K.: Analytics of an imperfect four-layer production inventory model under two-level credit period using branch-and-bound technique. J. Oper. Res. Soc. China 10, 725– 748 (2022). https://doi.org/10.1007/s40305-020-00300-1 Covert, D.P., Millan, J.A.O., Efendigil, T.: Dynamic customer service levels: evolving safety stock requirements for changing business needs. Enter. Bus. Manag. 15, 27–66 (2020). https://doi. org/10.5771/9783828872301-27 Freeland, J.R., Landel, R.: Managing inventories—reorder point systems. SSRN Electr. J. (2006). https://doi.org/10.2139/ssrn.911426 Etone, D.: The Establishment and Operation of the Universal Periodic Review, 1st edn. The Human Rights Council (2020)

Evaluating Research Impact: A Comprehensive Overview of Metrics and Online Databases Seema Ukidve1 , Ramsagar Yadav2(B) , Mukhdeep Singh Manshahia2 , and Jasleen Randhawa3 1 Department of Mathematics, L. S. Raheja College of Arts and Commerce, Santacruz(W),

Maharashtra, India 2 Department of Mathematics, Punjabi University Patiala, Patiala, Punjab, India

[email protected] 3 Panjab University Chandigarh, Chandigarh, India

Abstract. The purpose of this research paper is to analyze and compare the various research metrics and online databases used to evaluate the impact and quality of scientific publications. The study focuses on the most widely used research metrics, such as the h-index, the Impact Factor (IF), and the number of citations. Additionally, the paper explores various online databases, such as Web of Science, Scopus, and Google Scholar, that are utilized to access and analyze research metrics. The study found that the h-index and IF are the most commonly used metrics for evaluating the impact of a publication. However, it was also found that these metrics have limitations and cannot be used as the sole criteria for evaluating the quality of research. The study also highlights the need for a comprehensive and holistic approach to research evaluation that takes into account multiple factors such as collaboration, interdisciplinary work, and societal impact. The analysis of online databases showed that while Web of Science and Scopus are considered to be the most reliable sources of research metrics, they may not cover all relevant publications, particularly those in less well-established or interdisciplinary fields. Google Scholar, on the other hand, is more inclusive but may not have the same level of accuracy and reliability as the other databases. Keywords: Research metrics · Online databases · H-index · Impact factor · Citations · Web of science · Scopus · Google scholar · Research evaluation

1 Introduction Evaluating the impact of scientific research is crucial for researchers, institutions, and funding agencies to make informed decisions about the allocation of resources and the recognition of scientific achievements. In recent years, there has been a growing interest in developing methods and tools to measure and assess the impact of scientific research. One of the most commonly used methods is the analysis of citation data, which reflects the recognition and dissemination of scientific findings. In order to measure and assess the impact of scientific research, various research metrics and online databases have been developed. Research metrics are mathematical © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 236–243, 2024. https://doi.org/10.1007/978-3-031-50158-6_24

Evaluating Research Impact: A Comprehensive Overview of

237

formulas that are used to quantify the impact of scientific research based on citation data. Some of the most widely used research metrics include the H-index, g-index, eigenvectorbased citation index, etc. Online databases, such as Scopus, Web of Science, Google Scholar, Mendeley, and arXiv, provide a platform for tracking and analyzing scientific publications and citations. Despite the widespread use of research metrics and online databases for evaluating scientific impact, there are also limitations and biases associated with these tools. For example, some research metrics may not accurately reflect the impact of interdisciplinary or emerging fields, and some online databases may not have comprehensive coverage of all scientific disciplines. The aim of this research paper is to provide a comprehensive and objective overview of research metrics and online databases for evaluating scientific impact. The paper will examine the strengths and weaknesses of each research metric, compare and contrast the features and capabilities of online databases, and explore the limitations and biases associated with using these tools. The paper will also provide recommendations for researchers, institutions, and funding agencies for using research metrics and online databases effectively in evaluating the impact of scientific research.

2 Literature Review The H-index, a metric that measures both the productivity and impact of a researcher’s scientific publications. The H-index takes into account the number of publications and the number of citations received by each publication [1]. The H-index should be used in conjunction with other metrics to obtain a comprehensive picture of a researcher’s scientific impact [2]. The g-index takes into account the number of citations received by each article, rather than just the total number of citations received [3]. The eigenvectorbased citation index, a metric that measures the impact of a researcher’s publications by considering both the number of citations and the impact of the journals in which the publications appear [4]. The Science Citation Index Expanded and the Social Sciences Citation Index, two well-known databases that track scientific publications and citations. The author argues that these databases provide valuable information for evaluating the impact of scientific research, but should be used with caution due to potential biases and limitations [5]. Scopus is a database that provides information on scientific publications, including the number of citations received by each publication. It covers a wide range of scientific disciplines and is widely used for evaluating the impact of scientific research [6]. Web of Science is a database that provides information on scientific publications, including the number of citations received by each publication. It covers a wide range of scientific disciplines and is widely used for evaluating the impact of scientific research [7]. Google Scholar is a database that provides information on scientific publications, including the number of citations received by each publication. It covers a wide range of scientific disciplines and is widely used for evaluating the impact of scientific research [8]. Mendeley is a database and a reference management software that provides information on scientific publications, including the number of citations received by each publication. It covers a wide range of scientific disciplines and is widely used for organizing and evaluating the impact of scientific research [9].

238

S. Ukidve et al.

arXiv is a repository of electronic preprints of scientific articles in a variety of fields, including physics, mathematics, computer science, and biology. It allows researchers to share their work before it has been peer-reviewed or published in a journal, and provides a platform for the dissemination of new and innovative research [10].

3 Objectives The objectives of the research paper are as follows: • To provide an overview of various research metrics that are used to evaluate the impact of scientific research, such as the H-index, g-index, eigenvector-based citation index, etc. • To examine the strengths and weaknesses of each research metric and to identify the most appropriate metric(s) for different types of scientific research. • To present an overview of online databases that are used to track scientific publications and citations, including Scopus, Web of Science, Google Scholar, Mendeley, and arXiv. • To compare and contrast the features and capabilities of these online databases and to identify the most suitable database(s) for different types of scientific research. • To explore the limitations and biases associated with using research metrics and online databases to evaluate scientific impact, and to provide suggestions for overcoming these limitations. • To provide recommendations for researchers, institutions, and funding agencies for using research metrics and online databases effectively in evaluating the impact of scientific research.

4 Analysis and Comparison of the Various Research Metrics and Online Databases The evaluation of scientific publications is a complex process that involves multiple factors. Research metrics and online databases play a crucial role in measuring the impact and quality of scientific publications. In this paper, we will analyze and compare the various research metrics and online databases used to evaluate the impact and quality of scientific publications. One of the most widely used research metrics is the h-index, which measures both the productivity and citation impact of an author’s publications. The h-index provides a composite measure of an author’s research impact, taking into account both the number of publications and the number of citations received [11]. Another commonly used metric is the number of citations, which provides a measure of the impact of a publication based on the number of times it has been cited by other authors. The impact factor, which is calculated by dividing the number of citations in a given year by the number of articles published in the previous two years, provides an overall measure of the impact of a publication in a specific field [12]. Online databases such as Google Scholar, Web of Science, and Scopus provide a wealth of information about the research output of authors and institutions, and can be

Evaluating Research Impact: A Comprehensive Overview of

239

used to calculate these metrics. These databases also provide access to a large number of scientific articles, making it possible to compare the impact of different publications. It is important to note that these metrics have limitations, and it is essential to consider other factors when evaluating the impact and quality of scientific publications. For example, the quality of the research, the broader impact on society, and the interdisciplinary nature of the research should also be taken into account. A combination of metrics and databases is necessary to get a complete picture of the impact and quality of scientific publications. While no single metric or database can provide a definitive evaluation, using multiple sources of information can help provide a more comprehensive view of the impact and quality of scientific research.

5 Widely Used Research Metrics and the Number of Citations The h-index, Impact Factor (IF), and number of citations are some of the most widely used research metrics for evaluating the impact and quality of scientific publications. The h-index is a composite measure of an author’s research impact, taking into account both the number of publications and the number of citations received. It is calculated by ranking an author’s publications in terms of the number of citations they have received, and then selecting the highest number h such that h of the author’s papers have received at least h citations [13]. The Impact Factor (IF) is a metric that measures the average number of citations received by articles published in a specific journal. It is calculated by dividing the number of citations received by the number of articles published in the previous two years. The IF is widely used to evaluate the impact of a journal, and higher IFs are often seen as an indicator of higher quality research. The number of citations provides a measure of the impact of a publication based on the number of times it has been cited by other authors. It is widely used to evaluate the impact of an author or publication, with higher numbers of citations indicating higher impact [14]. It is important to note that these metrics have limitations, and should be used in conjunction with other factors, such as the quality of the research, the broader impact on society, and the interdisciplinary nature of the research, to provide a comprehensive evaluation of the impact and quality of scientific publications.

6 Various Online Databases that Are Utilized to Access and Analyze Research Metrics Web of Science, Scopus, and Google Scholar are some of the most widely used online databases for accessing and analyzing research metrics. These databases provide a wealth of information about the research output of authors and institutions, including information on the number of publications, citations, and other research metrics. Web of Science is a database that provides access to scientific literature in the natural sciences, social sciences, and humanities. It covers a broad range of disciplines and includes more than 33,000 journals, as well as conference proceedings, books, and other

240

S. Ukidve et al.

types of scientific literature. Web of Science provides detailed information on the number of citations received by each publication, as well as the h-index and Impact Factor of authors and journals. Scopus is a database that provides access to over 20,000 peer-reviewed journals, as well as conference proceedings, books, and other types of scientific literature. It covers a broad range of disciplines and provides detailed information on the number of citations received by each publication, as well as the h-index and Impact Factor of authors and journals. Scopus also provides access to a large number of citation reports, which provide a comprehensive overview of the research impact of authors and institutions. Google Scholar is a free online database that provides access to a large number of scientific articles, as well as conference proceedings, books, and other types of scientific literature. It covers a broad range of disciplines and provides information on the number of citations received by each publication, as well as the h-index of authors. Google Scholar also provides access to a large number of citation reports, which provide a comprehensive overview of the research impact of authors and institutions [15]. These online databases are widely used by researchers, institutions, and funding agencies to access and analyze research metrics, and to evaluate the impact and quality of scientific publications. They provide a valuable resource for understanding the contribution of research to the advancement of knowledge.

7 The Need for a Comprehensive and Holistic Approach to Research Evaluation Research evaluation is an important process for understanding the impact and quality of scientific publications, and for determining the allocation of research funding and recognition. However, relying solely on research metrics, such as the h-index, Impact Factor (IF), and number of citations, to evaluate research can be limited and can result in an incomplete picture of the impact of a publication. There is a growing recognition of the need for a comprehensive and holistic approach to research evaluation that takes into account multiple factors, such as collaboration, interdisciplinary work, and societal impact. Collaboration is becoming increasingly important in the scientific community, and it is important to consider the impact of collaborations in research evaluation. Interdisciplinary work is also becoming more common, and it is important to consider the interdisciplinary nature of research in evaluation, as this can have a significant impact on the impact of a publication [16]. Societal impact is another important factor that should be taken into account in research evaluation. Research that has a significant impact on society can be of great value, even if it may not have a high number of citations. It is therefore important to consider the broader impact of research on society when evaluating research. A comprehensive and holistic approach to research evaluation requires taking into account multiple factors and considering the interplay between different metrics. This can provide a more complete picture of the impact and quality of research, and can help ensure that research funding and recognition are allocated fairly and effectively. A comprehensive and holistic approach to research evaluation is essential for understanding the impact and quality of scientific publications, and for determining the allocation of research funding and recognition. This approach should take into account multiple

Evaluating Research Impact: A Comprehensive Overview of

241

factors, such as collaboration, interdisciplinary work, and societal impact, to provide a more complete picture of the impact of research.

8 The Need for Continuous Refinement and Improvement The use of research metrics, such as the h-index, Impact Factor (IF), and number of citations, to evaluate the impact and quality of scientific publications has been widely used for many years. However, as research becomes increasingly complex and multidimensional, it is important to continuously refine and improve these metrics to better reflect the nature of scientific research. Research is becoming increasingly interdisciplinary, with researchers working across multiple fields and disciplines. This can result in the production of highly impactful and innovative research, but it can also be difficult to measure and evaluate using traditional metrics. It is important to consider the interdisciplinary nature of research in the refinement and improvement of research metrics [17]. Additionally, research is becoming more global and collaborations are becoming increasingly common, leading to a more complex and diverse research landscape. The impact of research is not limited to the number of citations it receives, and other factors, such as societal impact, should also be considered in the evaluation of research [18]. Moreover, the rapid pace of technological development and the increasing availability of online platforms for research dissemination have changed the way research is conducted and evaluated. As a result, it is important to continuously refine and improve research metrics to better reflect the multidimensional nature of scientific research in the digital age [19]. The use of research metrics is a valuable tool for evaluating the impact and quality of scientific publications. However, as research becomes increasingly complex and multidimensional, it is important to continuously refine and improve these metrics to better reflect the nature of scientific research. This will help ensure that research funding and recognition are allocated fairly and effectively, and that the impact and quality of research are accurately reflected in the evaluation process [20].

9 Best Practices in Using Research Metrics and Online Databases A. Reliable Sources of Information • Using official databases and platforms such as Scopus, Web of Science, and Google Scholar • Avoiding sources that may not be reliable or may manipulate metrics B.. Comparison of Different Databases • Understanding the differences in data sources and coverage of each database • Using multiple databases to get a comprehensive view of an individual’s scientific impact C. Integration of Multiple Metrics • Using a combination of metrics, such as the h-index and g-index, to get a more nuanced view of an individual’s impact • Understanding the strengths and limitations of each metric

242

S. Ukidve et al.

D. Limiting Personal Biases • Being aware of potential biases in metrics and databases • Using metrics objectively, avoiding using them for personal gain or reputation building.

10 Conclusion Research metrics and online databases play a crucial role in evaluating the impact of scientific research. There are various research metrics, such as the H-index, g-index, and eigenvector-based citation index, that have been developed to quantify the impact of scientific research based on citation data. Online databases, including Scopus, Web of Science, Google Scholar, Mendeley, and arXiv, provide a platform for tracking and analyzing scientific publications and citations. However, it is important to recognize the limitations and biases associated with using research metrics and online databases to evaluate scientific impact. Some research metrics may not accurately reflect the impact of interdisciplinary or emerging fields, and some online databases may not have comprehensive coverage of all scientific disciplines. Despite these limitations, research metrics and online databases can provide valuable information for researchers, institutions, and funding agencies to make informed decisions about the allocation of resources and the recognition of scientific achievements. It is recommended that researchers, institutions, and funding agencies use research metrics and online databases effectively, taking into account the strengths and weaknesses of each tool and the limitations and biases associated with their use. Acknowledgements. Authors are grateful to Punjabi University, Patiala for providing adequate library and internet facility.

References 1. Hirsch, J.E.: An index to quantify an individual’s scientific research output. Proc. Natl. Acad. Sci. USA 102(46), 16569–16572 (2005) 2. Batista, P.D., Campiteli, M.G., Kinouchi, O.: Is it possible to compare researchers with different scientific interests? An analysis of the h-index. Scientometrics 68(3), 179–189 (2006) 3. Egghe, L.: Theory and practise of the g-index. Scientometrics 69(1), 131–152 (2006) 4. Radicchi, F., Fortunato, S., Castellano, C.: Universality of citation distributions: toward an objective measure of scientific impact. Proc. Natl. Acad. Sci. USA 105(45), 17268–17272 (2008) 5. Leydesdorff, L.: Mapping the global development of science by means of publication indicators: a study of Science Citation Index Expanded and Social Sciences Citation Index. J. Am. Soc. Inf. Sci. Technol. 61(7), 1386–1403 (2010) 6. Scopus (n.d.). https://www.elsevier.com/solutions/scopus 7. Web of Science (n.d.). https://clarivate.com/webofsciencegroup/ 8. Google Scholar (n.d.). https://scholar.google.com/ 9. Mendeley (n.d.). https://www.mendeley.com/ 10. arXiv (n.d.). https://arxiv.org/

Evaluating Research Impact: A Comprehensive Overview of

243

11. Bollen, J., Van de Sompel, H., Hagberg, A., Chute, R.: A principal component analysis of 39 scientific impact measures. PLoS ONE 4(6), e6022 (2009) 12. Bornmann, L., Leydesdorff, L.: What do citation counts measure? A review of studies on citing behavior. J. Document. 64(1), 45–80 (2008) 13. Garfield, E.: Citation analysis as a tool in journal evaluation. Science 214(4520), 671–681 (1979) 14. Hirsch, J.E.: Does the Hirsch index have predictive power? arXiv preprint arXiv:0707.3168 (2007) 15. Garfield, E.: Citation Indexing: Its Theory and Application in Science, Technology, and Humanities. Wiley, New York (1995) 16. Ioannidis, J.P.: Why most published research findings are false. PLoS Med. 2(8), e124 (2005) 17. Radicchi, F., Fortunato, S., Castellano, C.: Diffusion of scientific credits and the ranking of scientists. Phys. Rev. E 80(5), 056103 (2009) 18. Schreiber, M., Glassey, J.: A critical examination of the h-index in comparison with traditional indices and with peer judgement. Scientometrics 71(2), 317–331 (2007) 19. Van Raan, A.F.J.: Comparison of the Hirsch-index with standard bibliometric indicators and with peer judgment for 147 chemistry research groups. J. Am. Soc. Inform. Sci. Technol. 57(8), 1060–1071 (2006) 20. Waltman, L., van Eck, N.J.: A comparative study of four citation-based indices for ranking journals. J. Inform. 4(2), 146–157 (2010)

Hazard Identification, Risk Assessment and Control (HIRAC) at the Wood Processing Industry Herry Christian Palit(B)

and Alexander

Industrial Engineering Department, Petra Christian University, Siwalankerto 121-131, Surabaya, Indonesia [email protected]

Abstract. This study discusses the application of Hazard Identification, Risk Assessment, and control (HIRAC) at the wood processing industry which processes raw materials in the form of logs into various sizes of wood. This research was conducted due to the lack of awareness of the management concerning the occupational health and safety of its employees. It can be seen from the high number of accidents, where during 2016–2019 there were 16 cases. The results of the risk assessment using HIRAC showed 6% potential hazard with a low risk level, 29% with a moderate risk level, and 65% with a very high-risk level. Hazard control is focused on potential hazards with a very high-risk level. Hazard control consist of seven type of administrative controls, one type of elimination control, four types of engineering controls, and one type of personal protective equipment control. Keywords: Hazard identification · Risk assessment · Hazard control · Occupational health and safety

1 Introduction Occupational health and safety belong to the essential factors influencing towards labor’s productivity. In order to be able to work productively, working environment needs to be guaranteed safe and healthy. Hazard Identification and Risk Assessment is a process to define and identify hazard potentials; as well as to assess risk level of possible hazard occurrence by considering both probability and severity level. Every singly industry is required to identify potential hazard and assess its risk level within the process of establishing occupational safety and health guidelines. This risk assessment should be performed by utilizing the guidelines and standards of risk assessment [1]. Based on the proposed risk assessment, precise hazard prevention which eliminates and reduces hazardous potentials can be established. This research is performed in a wood processing industry in Surabaya cultivating timber as the raw material, into various wood products such as sawn timber, Slice Four Side (S4S), panels, decking, flooring, finger joint, and many others. So far, the industry’s © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 244–252, 2024. https://doi.org/10.1007/978-3-031-50158-6_25

Hazard Identification, Risk Assessment and Control (HIRAC) at

245

management has yet to possess consideration towards its labors’ safety and health, judging from the absence of reliable occupational hazard prevention and control. This is supported by the data on work-related accidents during 2016–2019, pointing out 16 accidents in which 15 of them are ergonomics-related (as displayed on Table 1). In addition, current working environment also potentially triggers occupational disease for the labors. This can be seen from wood dust scattered on the production floor area, which may trigger breathing and vision problems. Table 1. Data on work-related accidents Causing factors

Year 2016

2017

2018

2019

Physics

–

–

–

–

Chemics

1

–

–

–

Biology

–

–

–

–

Ergonomics

2

4

3

6

Psychics

–

–

–

–

Source Industrial database

Previous researchers have pointed out that HIRAC implementation in various industries brings various result on the risk hazard percentage level. The implementation of HIRAC in a boiler operation at Indonesian Power Unit, Semarang Ltd. Points out highrisk hazard percentage level in amount of 16.67% [2]. The implementation of HIRAC in the fabrication process at Pertamina Balongan Ltd. Ambarani and Tualeka [3] points out high risk hazard percentage level in amount of 45%, and very high-risk hazard percentage level at 5%. Indrawati et al. [4] points out 22% of high-risk hazard percentage level as a result of HIRAC implementation in a furniture industry. The better the implementation of occupational safety and health guidelines in an industry, the lesser its high and very high-risk hazard level percentage. This research is aimed to assist corresponding industries in identifying hazard potentials in a production floor, along with their risk level assessment. The higher the risk level, the more effort towards hazard control needs to be done, in order to reduce workrelated accidents and work-related disease among the labors. Thus, hazard prevention and control are focused on hazard potentials with high and very high-risk level.

2 Method Observation and HIRAC implementation are performed by involving both the owner and human resource manager. Their participation begins with identifying hazard potentials; determining probability and severity level for risk assessment; and hazard control on the production floor. There are 142 labors working on the production area, 106 males and 36 females, in which 6 are acting supervisors. They work approximately 40 h a week. During worktime, they are prone to the hazard potentials scattered around their working environment, both work-related accident and work-related disease.

246

H. C. Palit and Alexander

The initial stage for hazard identification is performed to identify process and activities that may trigger work-accident and work-related disease related problems. This is performed by observation utilizing ergonomic checkpoints from International Labor Office [5]. This tool is comprised of separate categories: Material storage and handling, which is intended to observe material storing and mobilization activity on the workspace; Hand tool, intended to observe activities related to the use of tools on the workspace; Machine safety, intended to observe the safety on the use of production machineries on the workspace; Workstation design, intended to observe both safety and convenience of the labors while working on their workspace; Lighting, intended to observe overall lighting and lighting intensity on the workspace; Premises, intended to observe air circulation on the workspace and evacuation system towards potential hazards such as fire and others; Hazardous substance and agents, intended to observe hazardous agents on the workspace; Welfare facilities, intended to observe general facilities; Work organization, intended to observe how the organization make decisions and policies. Hazard identification is afterwards continued by more detailed elaboration, in order to better identify hazard potentials and risks that may occur during activities on the production level area, both work-related accident and work-related disease. Based on the hazard identification stage, risk assessment is performed by considering the probability of the hazard, and the severity that may occur as an impact. Based on the multiplication process towards probability and severity scores, risk value is obtained. Table 2 presents Risks Assessment Matrix Model to analyze the category of risk level from each of the hazard potential. Table 2. Risk assessment matrix model Risk assessment matrix model

Probability

Severity Light injury (1)

Moderate injury (2)

Severe injury (3)

Fatality (4)

Disaster (5)

Very high (5)

Low

Moderate

High

Very high

Very high

High (4)

Low

Moderate

High

Very high

Very high

Moderate (3)

Low

Moderate

Moderate

Tinggi

Tinggi

Low (2)

Low

Low

Moderate

Moderate

Moderate

Very low (1)

Low

Low

Low

Low

Low

Source Rout and Sikdar [6]

Hazard control mechanism in this research is focused towards hazard potentials with high and very high risk factor. According to the policy established by the Ministry of Labor number 5, 2018 [7], hazard control on the working environment should be performed according to the five hierarchical levels as pointed out by Fig. 1.

Hazard Identification, Risk Assessment and Control (HIRAC) at

247

Fig. 1. The hierarchy of workplace environment hazard control

Hazard control in a working environment is performed by conforming priority levels set on the Figure. The first hierarchy, or control through hazard elimination is a first priority. If not possible, then second layer of the control can be utilized.

3 Results and Discussion Hazard identification is performed by analyzing both of the result of field observation and measurement on the production area. Based on the overall wood processing, 21 activities with hazardous potential are identified. The following are examples list hazard potential during timber receiving, trimming, and maintaining stages. Mobilizing timber from truck with forklift. This belongs to the receiving and mobilizing stage. Hazard potentials found during this stage include forklift path, which appears to be the same one walked by the labors during their activity. This unorganized route may potentially cause accident, such as a worker may be hit by the forklift. Operating log trim machinery. This belongs to the trimming process, in which timbers are trimmed by using machineries. These machineries possess hazard potential for the operators, such as amputation, pinching, and direct exposure from the wood dust, noise, hot airflow, and high level of humidity. Machineries check and repair. This belongs to one of the activities during maintaining stage. Hazard potential in this activity may occur if the machineries suddenly turned on during a repair process. Risk assessment is performed after hazard identification by utilizing matrix displayed on Table 2. This aims to determine policy priority regarding problems or potential problems. This analysis is divided into two types of risks: work-related accident and work-related disease. Risk assessment is conducted by determining opportunity value and severity rate towards existing hazard potential. Probability assessment and severity risk level involve both owner and human resource manager. The probability of hazard potential related with work-related accident and work-related disease is determined through some values representing the likelihood an accident may occur. This probability is divided into 5 different levels, starting from the lowest level of probability. The description of each level can be seen on Table 3 (for work-related accident) and Table 4 (for work-related disease).

248

H. C. Palit and Alexander Table 3. Work-related accident probability level description

Level Description 1

Never happened/heard on the equivalent industries around the world

2

Ever happened/heard on the equivalent industries around the world

3

Ever happened once or more on the late three years, on the equivalent industries around the world

4

Ever happened once or more on the late three years, on the equivalent industries in Indonesia

5

Ever happened once or more on the late three years, on the industry

Source Adaptation from Ardani et al.[8]

Table 4. Work-related disease probability level description Level

Description

1

Rare exposure and low likelihood of causing disease

2

Low exposure and low likelihood of causing disease

3

Regular/irregular exposure and moderate likelihood of causing disease

4

Frequent exposure and high likelihood of causing disease

5

Constant exposure dan and very high likelihood of causing disease

Source Adaptation from Ambarani and Tualeka[3]

Severity level is assessed in order to find out the impact that may occur if hazard potential turns into realization. This severity level is divided into 5 different levels starting from the lowest severity level. The description of each level can be seen on Table 5 (for work-related accident) and Table 6 (for work-related disease). The result of probability and severity level is used to determine risk levels, which categorized from the lowest, as pointed out from Table 2. The followings are risk assessments towards hazard potential belonging to very high-risk category. Mobilizing timbers from a truck, mobilizing timbers on the production floor, and mobilizing timbers to the production storage may inflict accident between forklift and labors. The accident may cause broken bones and death; thus, severity value is given 5. The accident probability is also given 5 considering the fact that such accident happened once in 2019. Based on both severity and probability level, forklift accident is categorized as very high risk. Operators performing activity around trimming machine may potentially slipped and pinched by a moving wheeled cart. This may result in broken bones or death; thus the severity level is given 5. The probability of accident is also given 5 considering the fact that such accident happened once in 2019. Based on both severity and probability level, wheeled cart accident is categorized as very high risk.

Hazard Identification, Risk Assessment and Control (HIRAC) at

249

Table 5. Work-related accident severity level description Level Description 1

Near miss requires documentation and action

2

Potentially causing injury which require first aid kit

3

Require medical treatment, but does not cause work restriction or worktime lost

4

One or more injuries requiring medical treatment, which cause worktime lost, but does not inflict permanent damage

5

Potentially causing permanent damage or death

Source Adaptation from Ardani et al. [8]

Table 6. Work-related disease severity level description Level Description 1

Exposure requires documentation and action

2

Potentially causing disease which require first aid kit

3

Potentially causing disease, but does not cause work restriction or worktime lost

4

Potentially causing temporary disease, but require medical treatment which impacts on worktime lost

5

Potentially causing permanent damage or death

Source Adaptation from Ambarani and Tualeka [3]

The operator of trimming machinery may constantly exposed from wood dust, as he would always be on the same station during his worktime. The possibility of work-related disease is therefore high and given value of 5. Wood dust may cause eyes irritation, occupational asthma, or other chronical obstructive on the ones exposed, therefore severity level is also given 5. Considering such probability and severity level, this exposure is categorized as very high-risk category. Machinery turned on during maintenance/repairis considered very dangerous and may cause fatality on the victim, such as scratch or dismembered body parts, therefore severity level is given 5. The probability value is also given 5, considering the fact that such accident happened once in 2019. Considering such probability and severity level, this malfunctioning is categorized as very high risk. The result of assessment towards all activities within production area points out three types of risk categories: high risk in amount of 65%, moderate risk in amount of 29%, and low risk in amount of 6%, as displayed in Fig. 2. This implies that hazard control will be prioritized towards activities containing very high-level risk. This hazard control aims to reduce risk level and make risks more acceptable. Hazard control is performed towards very high-risk activities. The followings are how hazard control may be implemented towards very high-risk activities.

250

H. C. Palit and Alexander

Fig. 2. Working-environment potential risk category diagram

Establishing different path for transportation access and labors access. This control is performed to overcome problems on timber receiving process, temporary storage, and finished products storage. These three processes contain similar activities which is the mobilization of timbers by forklift. This control belongs to technical manipulation category, as the creation of specific path according to the purpose ensure the safety of labors walking on the path. Labors path can be created by applying paint with a-meter wide size in one of the sections of the transportation path. This paint functions to mark the area that can be specifically passed through by labor only. Other transportation devices such as forklift is not allowed to pass through it. Likewise, this control also forbids the labor to walk on the transport path. Labelling caution sign and designate safe area for the operator. This control aims to overcome problems that may happen during trimming process. Labelling ‘caution’ sign indicating the danger may prevent labors to walk too close with wheeled cart, therefore reducing pinching accident. This label can be installed on the outer-lower side of the cart. The exact location of the label can be determined by sticking out a sticker in the area. This prevention belongs to administrative category, as it helps labors to avoid area with high level of hazard. Installing wooden barrier on the surrounding of cart track can also support the prevention effort. This belongs to technical manipulation category, as this limits operator movement from slipping backwards. Installing local exhaust in production machineries and wearing protection devices. The installation of local exhaust belongs to technical manipulation category, as it prevents labors from being exposed to wood dust resulted from the production machineries. Wood dust may inflict eyes irritation, occupational asthma, and chronic diseases for those exposed. Local exhaust can be made in form of a pipe sucking scattered wood dust. The pipe should be pointed to the trimming location where wood dust appears. This prevents wood dust to reach labors’ breathing zone (300 mm diameter) [9]. This control can also be achieved by wearing mask and safety google glass. The aforementioned mask is quarter-mask-air-purifying respirator type, which equipped with dust filter made of fiberglass. Safety google glass prevents operator’s eyes from the exposure of wood dust, therefore preventing eyes irritation. The application of both mask and safety google glass should be made compulsories for all the workers in the production area. This conforms with the industry’s regulation regarding the use of protective devices.

Hazard Identification, Risk Assessment and Control (HIRAC) at

251

Providing label and marking area with potential hazard (hand caught inside a machine). This is performed especially in wood processing process involving the use of machineries. Providing label or danger sign on the machineries utilized on this stage is aimed to prevent accident caused by labors’ hand getting pulled by one of the machines. This control is performed by marking areas forbidden to be touched by the labors’ hands. Marking is done by sticking out stickers showing danger sign on the area. This belongs to administrative category, as it helps the labors to avoid areas potentially hazardous towards their body part (in this case, hand getting caught inside a machine). Locking machine from the power source by using Lock Out/Tag Out (LOTO). LOTO system is a combined system aimed to prevent a machine from turning on during maintenance process. Lock out system locks the power source of a machine to prevent sudden activation, while tag out helps other labors in the vicinity to be aware of an ongoing maintenance process. LOTO system belongs to administrative control category, as labors are required to apply the system first, to ensure that the corresponding machine is isolated from the power source. This can be done by switching off electrical panel of the machine, followed by locking it with a padlock equipped with label explaining ongoing maintenance process. This deactivated and isolated power unit prevents other unaware labors from switching on the machine during maintenance. For the future, the company is suggested to consider utilizing technology in order to better manage the hazard control. The utilization of safety sensor installed on the machineries and hazardous areas potentially be the hazard control [10, 11]. One of the applications of safety sensor on machineries is proximity sensor. The way this sensor works may vary depending on the transmission of light, ultrasonic, or other kind of waves. In general, proximity sensor transmits wave and calculate its wavelength according to the preset range. If it detects an object, the length of wave received back by the sensor will be different as it is bounced by the object. If any object starts entering hazardous area with conveyor, vacuum, or anything similar, integrated proximity sensor will release a particular warning. Sound or machine that turns off automatically will be made on at the moment proximity sensor detects an object in proximity. Passive Infrared Receiver (PIR) sensor detects any foreign object being too close with the hazardous area will release a loud sound, as a warning towards the object. In addition, the use of wearable technologies to observe workers in real-time and to remove potential hazard is also important to consider. Hazard originated from wood dust and machinery noise can be monitored by wearable technology. The NIOSH Center for Direct Reading and Sensor Technologies (NCDRST) has developed several wearable technologies to improve occupational health and safety [12]. Respirable Dust Sensor Technology can be used to monitor the exposure of wood dust, whether it is still safe or it has trespassed the threshold. Sound level meter apps, utilizing embedded smartphone sensors is a tool to measure noise level in a workplace. This should be supported with noise exposure perimeter to better reduce hearing loss caused by noise in a workplace. With the help of technology, occupational health and safety management is expected to better provide hazard-free environment more effectively in a workplace.

252

H. C. Palit and Alexander

4 Conclusion The result of this research points out that working environment in the production area potentially harms labors’ health and triggers occupational hazards. This can be seen from the number of very high-risk level of activities in amount of 65%, while the 29% is moderate risk level activities, and the remaining 6% as low risk level activities. The activities with very high level of risk becoming the subject priority of hazard control. Proposed hazard control includes seven administrative control category, one elimination control category, four technical manipulation category, and one protective device control category.

References 1. Ramesh, R., Prabu, M., Magibalan, S., Senthilkumar, P.: Hazard identification and risk assessment in automotive industry. Int. J. ChemTech Res. 10(4), 352–358 (2017) 2. Zeinda, E.M., Hidayat, S.: Risk assessment kecelakaan kerja pada pengoperasian boiler di PT. Indonesia Power Unit Pembangkitan Semarang. Indones. J. Occup. Saf. Health 5(2), 183–191 (2016) 3. Ambarani, A.Y., Tualeka, A.R.: Hazard identification and risk assessment (HIRA) pada proses fabrikasi plate tanki 42-T-501A PT. Pertamina (PERSERO) RU VI Balongan. Indones. J. Occup. Saf. Health 5(2), 192–203 (2016) 4. Indrawati, S., Prabaswari, A.D., Fitriyanto, M.A.: Risk control analysis of a furniture production activities using hazard identification and risk assessment method. MATEC Web Conf. 154, 01102 (2018). https://doi.org/10.1051/matecconf/201815401102 5. International Labour Office: Ergonomic Checkpoints: Practical and Easy-to-Implement Solutions for Improving Safety, Health and Working Conditions. International Labour Office, Switzerland (2010) 6. Rout, B.K., Sikdar, B.K.: Hazard identification, risk assessment, and control measures as an effective tool of occupational health assessment of hazardous process in an iron ore pelletizing industry. Indian J. Occup. Environ. Med. 21(2), 56–76 (2017). https://doi.org/10.4103/ijoem. IJOEM_19_16 7. Kemenaker: Peraturan Menteri Ketenagakerjaan Republik Indonesia (PERMENAKER) nomor 5 tahun 2018 tentang Keselamatan dan Kesehatan Kerja Lingkungan Kerja. Kementerian Ketenagakerjaan Republik Indonesia, Jakarta (2018) 8. Ardani, H.N., Santoso, H., Rumita, R.: Analisis risiko kesehatan dan keselamatan kerja pada pekerja divisi mill boiler (Studi Kasus di PT Laju Perdana Indah PG Pakis Baru, Pati). Ind. Eng. Online J. 3(2), 1–6 (2014) 9. Harrianto, R.: Buku Ajar Kesehatan Kerja. Penerbit Buku Kedokteran EGC, Jakarta (2009) 10. Rajendran, S., Giridhar, S., Chaudhari, S., Gupta, P.: Technological advancements in occupational health and safety. Meas. Sens. 15, 100045 (2021). https://doi.org/10.1016/j.measen. 2021.100045 11. Haas, E.J., Cauda, E.: Using core elements of health and safety management systems to support worker well-being during technology integration. Int. J. Environ. Res. Public Health 19(21), 13849 (2022). https://doi.org/10.3390/ijerph192113849 12. The National Institute for Occupational Safety and Health (NIOSH). https://www.cdc.gov/ niosh/programs/cdrst/default.html

Location Selection of Rail Transportation Hub Using TOPSIS Method Kanokporn Sripathomswat(B) , Nattawat Tipchareon, Worapat Aruntippaitoon, Itiphong Trirattanasarana, and Sunarin Chanta King Mongkut’s University of Technology North Bangkok, 1518, Pracharat 1 Road,Wongsawang, Bangsue, Bangkok 10800, Thailand [email protected], [email protected]

Abstract. Nowadays, the transportation system has a significant impact on the economic drive, especially in densely populated areas. The rapid development of the transportation system has resulted in people having several transportation options. However, people still have problems accessing the origin station and transiting from the terminal station to their destination. The purpose of this study is to select the best location of the transportation hub for the Red Line suburban train, Bang Sue—Rangsit route, Thailand. Since the problem involves several factors, the multiple criteria decision-making method, TOPSIS, is selected to solve the problem. We select the criteria based on the main factors, which are connectivity, accessibility, the number of passengers, and the potential for station expansion. Based on the concept of TOPSIS, the positive ideal value of each criterion is identified, then the relative closeness score of each alternative is calculated. The best location of the transportation hub of the Red Line is the station with the highest relative closeness score. The results provide a decision support model and help decision-makers in designing transportation hub location planning. Keywords: Railway · Transportation hub · Public transportation · Multiple criteria decision-making · TOPSIS

1 Introduction The development of rail transportation is an important mechanism for driving the economy and developing the country. However, the development of rail transportation requires a very high investment budget. Entrepreneurs therefore need to plan for longterm returns. In the early stages of enabling the rail transport system, many times there is a loss problem, because consumers are familiar with the traditional transportation system used to travel and inconvenient to change travel behavior. Moreover, rail transport cannot provide a door-to-door service. It requires connecting a bus/taxi to the departure station or from the terminal station to the destination, which causes inconvenience for Please note that the AISC Editorial assumes that all authors have used the western naming convention, with given names preceding surnames. This determines the structure of the names in the running heads and the author index. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 253–261, 2024. https://doi.org/10.1007/978-3-031-50158-6_26

254

K. Sripathomswat et al.

travel. Sometimes it takes more total travel time from the origin to the destination or incurs more total travel expenses compared to traveling by road transport. In the early phase of rail service operations, the planning of travel connection points may not yet cover the needs of consumers. Especially, for a large station that sever a massive number of passengers, it will cause long waiting times for connecting buses/taxis. In this research, we present a guideline for analyzing the optimal location of the transportation hub or connecting points for the new Red Line suburban train, Bang Sue—Rang Sit route, Bangkok, Thailand. The objective is to find the best location for a railway transportation hub, which will be used as a connecting/transit point to other transportation modes in order to provide traveling convenience for rail passengers. In this case, not only the number of passengers that matter, but also other related factors are taken into account such as connectivity, physical characteristics of the train station, the available transportation modes and transits around the station, and the number of government offices and business centers around, etc. Since the decision concerns many factors, the multiple criteria decision-making method is selected as a tool for solving the problem. The Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is selected for deciding the best location of a rail suburban transportation hub. The important level of the criteria is evaluated by railway experts. The results can be used as a guideline for decision-makers in hub location planning.

2 Literature Review Designing a public transport network system involves many components and factors. The main structural components of a transport network are nodes, links, hubs, and flows [1]. Node is defined as a location that has access to the transportation network, whereas link is a transport that is connected between two nodes. Flow is the amount of traffic that is transported on a link between two nodes. Hub is a node that handles the amount of flow and connects transit for different transportation modes. In this study, we focus on transportation hub selection problems, where the system performance depends on hub location. Several works have been proposed for transport hub network design [2–5]. Nielsen and Lange [6] listed the success factors for transportation network design. Since determining the appropriate location of the hub requires many factors to be considered, multiple criteria decision-making (MCDM) methods are used to solve the problems. Wang et al. [7] proposed an evaluation scheme based on TOPSIS including AHP and Entropy for the city rail transit network. They selected 11 indices such as network scale, passenger flow, ticket revenue, etc. from an evaluation of 3 aspects: network function, traffic function, and economics. The evaluation matrix was standardized by [0,1] linear transformation. Weights of influence parameters were determined by AHP and entropy, considered expert knowledge and experience. The weighted standardized decision-making matrix was built up through a standardized decision-making matrix and parameter weights. The selection of highway network planning schemes was determined by calculating the close degree of the schemes and the ideal schemes based on the TOPSIS method. Zabihi et al. [8] considered the best location for a container transshipment hub in the southern seas of Iran using a hybrid MCDM method based on Analytical Hierarchy

Location Selection of Rail Transportation Hub Using TOPSIS Method

255

Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The problem was considered as three levels: goal, criteria, and alternatives. To select the best location for a transshipment hub, six main criteria were considered, which are port location, port physical, hinterland economy, port efficiency, cost, and other condition. The criteria that have the highest weight was the port location, which composited of three sub-criteria: the closeness to the import/export area, proximity to the feeder port, and the closeness to navigation routes. A total of 18 sub-criteria were evaluated as positive and negative dimensions using the distance of the positive-ideal and negative-ideal solutions, respectively. The final ranking showed that Bandar Abbas port is the best location for a transshipment hub among 6 alternatives. Chen et al. [9] proposed a selection model for logistic centers based on TOPSIS and Multichoice Goal Programming (MCGP) methods. They integrated fuzzy technique for order preference by similarity to an ideal solution and MCGP to obtain an appropriate logistics center from many alternative locations for the airline industry. They selected five criteria composited of both qualitative and quantitative data, which are resource availability, location resistance, expansion possibility, investment cost, and information abilities. The decision-making group (DM) was required to select the best logistic center from 5 candidates by applying the Delphi technique. The importance fuzzy weights of the criteria were determined by the DM. The fuzzy positive-ideal and fuzzy negative-ideal were determined, then the distance of each location candidate from fuzzy positive-ideal and fuzzy negative-ideal was calculated with respect to each criterion. The closeness coefficients obtained for each location candidate were used as priority values to build the TOPSIS-MCGP model. Zhao et al. [10] determined a location of intra-city distribution hub in the urban metro network using a segmentation method. The indices for evaluating the importance of each metro station were selected by complex network theory. The evaluation index weight was calculated by AHP method, then the importance of each metro station was evaluated using the TOPSIS model. The location model was formulated to find the final metro distribution hubs from the candidate metro distribution hubs with consideration of logistics demand. They chose the Shanghai metro system as a case study and provided strategic planning for metro distribution hub location selection.

3 Material and Methods 3.1 Multiple Criteria Decision-Making Methods MCDM is a complex decision-making tool that involves both quantitative and qualitative factors [11]. MCDM is a branch of operations research for solving multiple criteria problems. MCDM methods are classified into two categories: discrete MCDM or Multiple Attribute Decision Making (MADM) and continuous Multiple Objective Decision Making (MODM) [12, 13]. In this study, we focus on the MADM. The MADM methods include TOPSIS, AHP, Simple Additive Weighting (SAW), Weighted Product Method (WPM), Elimination and Choice Expressing Reality (ELECTRE), and Linear Programming Technique for Multidimensional Analysis of Preference (LINMAP), etc.

256

K. Sripathomswat et al.

3.2 Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) TOPSIS is one of several multi-criteria decision analysis methods to determine the best alternative. The TOPSIS method is to try to find an alternative whose overall performance is close to the best value for each criterion and far from the worst in each criterion as well. By applying TOPSIS, the analyst must make giving importance to each decision-making criterion. TOPSIS is a ranking tool among choices based on evaluation data of each choice of multiple-criteria decision-making. Therefore, it is suitable for making decisions using quantitative criteria that can be evaluated. The choice came out as incandescent numbers. The TOPSIS data analysis process follows the next step. Step 1: Create a decision matrix G that consists of m alternatives and n criteria, as shows in Eq. (1). Bij is the attribute value of the i alternative with respect to the j criteria, where i = 1, 2, …, n, j = 1, 2, …, m, n = number of alternatives, m = number of criteria, Ai = Alternative i, Bj = Criteria j. Step 2: Normalize the decision matrix to adjust the value of each criterion to be in the same standard. The elements of the normalized decision matrix r ij are given by Eq. (2). The value should be in the range of 0–1, and the total weights of all values must be equal to 1.

A1 A2 G= . .. Am

B1 B2 . . . Bn ⎤ . b11 b12 . . b1n ⎢ b21 b22 . . . b2n ⎥ ⎢ ⎥ ⎢ . ⎥ .. ⎣ .. ⎦ . ⎡

(1)

bm1 bm2 . . . bmn

rij = 

bij m i=1

; ∀i, j

(2)

b2ij

Step 3: Calculate the weighted normalized decision matrix. Let vij be the element of the weighted decision matrix, so it can be calculated by using Eq. (3) wj is the weight of criterion j, and the element of a vector of weights W that consist of n criteria, where the sum of the elements equal to 1, W = (w1 , w2 , …, wn ). vij = wj rij ; ∀i, j

(3)

Step 4: Determine the positive ideal value and the negative ideal value, where S + is the best value in each criterion when comparing all available alternatives, and S − is the worst value for each criterion when comparing all available alternatives. The 



(4) S + = v1+ , v2+ , . . . , vn+ = max vij i



 −

S − = v1− , v2− , . . . , vn

 = min vij i

(5)

Step 5: Compute the distance of each alternative from the positive ideal and the negative ideal solutions. Let d i + and d i − represent the distances of the ith alternative

Location Selection of Rail Transportation Hub Using TOPSIS Method

257

from the positive ideal solution and the negative ideal solution, respectively. Di and di can computed as in Eqs. (6)–(7)

 2

n  + Vij − Vj+ ; ∀i di =  (6) j=1

di−



 2

n  Vij − V − ; ∀i = j

(7)

j=1

Step 6: Obtain the relative closeness of each alternative to the ideal solution, where the relative closeness of the ith alternative to the ideal solution is defined as in Eq. (8). The highest C i indicates the best alternative. Ci∗ =

di−

di− + di+

; ∀i

(8)

3.3 A Case Study of Red Line Rail, Bangkok, Thailand The Red Line Rail is the first phase of the suburban railway system project that operates between Bang Sue Grand Station to Rangsit Station, Thailand. The Red Line sky train starts operating service in August 2021. The line is located in the area of 2 provinces, which are Bangkok (Dusit District, Phaya Thai District, Bang Sue District, Chatuchak District, Lak Si District and Don Mueang District) and Pathum Thani Province (Muang District, Thanyaburi District, Lam Luk Ka District and Khlong Luang District). There are 10 railway stations in the first phase, which compose of Bang Sue, Chatuchak, Wat Samian Nari, Bang Khen, Thung Song Hong, Lak Si, Kan Kheha, Don Mueang, Lak Hok, Rangsit. Figure 1 shows the connection of all 10 stations. After reviewing the literature related to factors that should be considered for designing transportation hubs, the factors are selected based on the urban planning concept in Bangkok [14]. We defined 10 criteria for making the decision on selecting the appropriate transportation hub for the Red Line Rail: (1) the distance from the station to other transportation mode transits (2) the variety of transportation system around the station (3) the number of exits of station (4) size of the station (5) the number of floors or the number of levels at the station (6) the number of transportation service points around the station (7) the number of passenger pick-up points below the station (8) area of car parking for passengers (9) the number of government offices and business centers around the station (10) the number of passengers at the station. Then each criterion is weighted by experts’ expertise. We rearrange all criteria by their important scores using Rank Order Centroid method (ROC) [11] as the following: (C1) the number of passengers at the station, (C2) the distance from the station to other transportation mode transits, (C3) the number of government offices and business centers around the station, (C4) the variety of transportation systems around the station, (C5) the number of transportation service points around the station, (C6) area of car parking for passengers, (C7) the number of passenger pick-up points below the station, (C8) the number of exits of station, (C9) size of station, (C10) the number of floors or the number of levels at the station.

258

K. Sripathomswat et al.

Fig. 1. The stations in Red Line Rail, Bangkok, Thailand

4 Results and Discussion In this case, n represents the number of alternatives (stations), where n = 10, and m represents the number of criteria (factors), where m = 10. The matrix of evaluated value of the 10 criteria for the 10 stations as shown in Table 1, where C1, …, C10 represented criteria 1–10, respectively. Based on the TOPSIS procedure in the previous section, we create a decision matrix G as shown in Table 2. Then, we calculated the normalized weighted decision matrix as shown in Table 3. The positive ideal values (S + ) and negative ideal values (S − ) of each criterion are calculated and reported in Table 4. Next, we calculate the relative closeness (C + ) and ranking of each alternative as shown in Table 5. The best location for building a transportation hub is defined as the highest rank of the station that has the highest relative closeness value. In this case, the Rung Sit station is the first rank with the highest relative closeness score of 0.724.

5 Conclusion and Future Work In this study, we consider the problem of selecting an appropriate railway transportation hub using the MCDM method. Designing a transportation network involves many factors. TOPSIS is selected to handle this problem. A case study of the sky train, Red Line, in Thailand is presented. There are 10 alternatives on the Red Line to be considered as a hub. Ten important factors are selected as criteria for selecting the appropriate hub for the

Location Selection of Rail Transportation Hub Using TOPSIS Method

259

Table 1. Evaluated value matrix. C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

Station 1

9

6.33

3

9

9

9

3

5

9

5

Station 2

1

3

1

3

3

5

3

5

3

5

Station 3

3

5

3

3

5

3

9

7

3

5

Station 4

3

3.83

7

3

5

3

7

7

3

5

Station 5

1

7

3

5

3

3

5

7

3

5

Station 6

5

6.83

9

7

9

1

7

7

3

5

Station 7

1

4.59

1

7

3

3

5

7

7

5

Station 8

9

4.83

3

9

9

7

7

5

3

5

Station 9

3

1.16

1

1

1

1

3

3

3

3

Station 10

9

6

5

9

3

3

3

3

7

5

ROC weight

0.293

0.193

0.143

0.109

0.085

0.065

0.048

0.033

0.021

0.010

Table 2. Weighted decision matrix. C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

Station 1

0.521

0.388

0.215

0.453

0.495

0.633

0.169

0.272

0.579

0.327

Station 2

0.058

0.184

0.072

0.151

0.165

0.352

0.169

0.272

0.193

0.327

Station 3

0.174

0.307

0.215

0.151

0.275

0.211

0.051

0.381

0.193

0.327

Station 4

0.174

0.235

0.503

0.151

0.275

0.211

0.395

0.381

0.193

0.327

Station 5

0.058

0.429

0.215

0.252

0.165

0.211

0.282

0.381

0.193

0.327

Station 6

0.290

0.419

0.646

0.353

0.495

0.070

0.395

0.381

0.193

0.327

Station 7

0.058

0.281

0.072

0.353

0.165

0.211

0.282

0.381

0.450

0.327

Station 8

0.521

0.296

0.215

0.453

0.495

0.493

0.395

0.272

0.193

0.327

Station 9

0.174

0.071

0.072

0.050

0.055

0.071

0.169

0.163

0.193

0.196

Station 10

0.521

0.368

0.359

0.453

0.165

0.211

0.169

0.163

0.450

0.327

railway line, which include the physical characteristics of the train station, the number of passengers, the available transportation modes and transits around the station, and the number of government offices and business centers around the station. All factors are assigned important weighted based on the railway expertise. Then, the decision matrix, which consists of the attribute value with respect to the particular alternative and criteria is created. Then, the decision matrix is normalized, and the positive and negative ideal solutions of each criterion are determined. The distances from the positive and negative ideal values and negative ideal values are calculated, and the relative closeness of each alternative is obtained. Finally, the best alternative for selecting a railway transportation

260

K. Sripathomswat et al. Table 3. Normalized weighted decision matrix. C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

Station 1

0.153

0.075

0.031

0.049

0.042

0.041

0.008

0.009

0.012

0.003

Station 2

0.017

0.036

0.010

0.017

0.014

0.023

0.008

0.009

0.004

0.003

Station 3

0.051

0.059

0.031

0.017

0.023

0.014

0.024

0.013

0.004

0.003

Station 4

0.051

0.045

0.072

0.017

0.023

0.014

0.019

0.013

0.004

0.003

Station 5

0.017

0.083

0.031

0.028

0.014

0.014

0.014

0.013

0.004

0.003

Station 6

0.085

0.081

0.092

0.038

0.042

0.005

0.019

0.013

0.004

0.003

Station 7

0.017

0.054

0.010

0.038

0.014

0.014

0.014

0.013

0.009

0.003

Station 8

0.153

0.057

0.031

0.049

0.042

0.032

0.019

0.009

0.004

0.003

Station 9

0.051

0.014

0.010

0.006

0.005

0.005

0.008

0.005

0.004

0.002

Station 10

0.153

0.071

0.051

0.049

0.014

0.014

0.008

0.005

0.009

0.003

Table 4. Positive ideal values and negative ideal values. C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

S+

0.152

0.082

0.092

0.049

0.042

0.041

0.021

0.012

0.012

0.003

S−

0.017

0.013

0.010

0.005

0.004

0.004

0.008

0.005

0.004

0.002

Table 5. Relative closeness and ranking of alternatives. ST1

ST2

ST3

ST4

ST5

ST6

ST7

ST8

ST9

ST10

C+

0.720

0.156

0.340

0.404

0.330

0.633

0.247

0.698

0.171

0.724

Ranking

2

10

6

5

7

4

8

3

9

1

hub is identified as the highest rank of relative closeness. The results can be used as a guideline for hub location planning. In this study, we consider the problem based on the viewpoint of operators. For future work, we plan to consider the problem into 3 facets: customers, operators, and providers. The MODM methods will be applied for solving the problem with 3 objectives based on 3 stakeholders who involve in the railway transportation system. So, we can have a practical solution from a broader point of view. Acknowledgments. This research was funded by National Science, Research and Innovation Fund (NSRF), and King Mongkut’s University of Technology North Bangkok with Contract no. KMUTNB-FF-66-67.

Location Selection of Rail Transportation Hub Using TOPSIS Method

261

References 1. Rodrigue, J.: Structural Components of Transport Networks. The Geography of Transportation System, Dept. of Global Studies and Geography, Hofstra University, New York (2023). https://transportgeography.org/contents/chapter2/geography-of-transportationnetworks/transport-network-structural-components/ 2. Campbell, J.F.: Integer programming formulations of discrete hub location problems. Eur. J. Oper. Res. 72, 387–405 (1994) 3. Hwang, Y.H., Lee, Y.H.: Uncapacitated single allocation p-hub maximal covering problem. Comput. Ind. Eng. 63, 382–389 (2012) 4. Peker, M., Kara, B.: The p-hub maximal covering problem and extensions for gradual decay functions. Omega 54, 158–172 (2015) 5. Farahani, R.Z., Hekmatfar, M., Arabani, A.B., Nikbakhsh, E.: Hub location problem: a review of models, classification, solution techniques, and applications. Comput. Ind. Eng. 64, 1096– 1109 (2013) 6. Nielsen, G., Lange, T.: Network Design for Public Transport Success: Theory and Examples (2007). https://thredbo-conference-series.org/downloads/thredbo10_papers/thredbo10themeE-Nielsen-Lange.pdf 7. Wang, L., Chen H., L., X.W.: City rail transit network schemes evaluation based on TOPSIS, including AHP and entropy. In: Proceedings of the 12th COTA International Conference of Transportation Professional, August 3–6, pp. 1731–1738. Beijing, China (2012) 8. Zabihi, A., Gharakhani, M., Afshinfar, A.: A multi criteria decision-making model for selecting hub port for Iranian marine industry. Uncert. Supply Chain Manag. 4, 195–206 (2016) 9. Chen, K., Liao, C., Wu, L.: A selection model to logistic centers based on TOPSIS and MCGP methods: the case of airline industry. J. Appl. Math. 157, 470128 (2014) 10. Zhao, L., Li, H., Li, M., Sun, Y., Hu, Q., Mao, S.: Location selection of intra-city distribution hubs in the metro-integrated logistics system. Tunnel. Undergr. Space Technol. 80, 246–256 (2018) 11. Mardani, A., Jusoh, A., Nor, K., Khalifah, Z., Zakwan, N., Valipour, A.: Multiple criteria decision-making techniques and their applications: a review of the literature from 2000 to 2014. Econ. Res. Ekonomska Istraživanja 28(1), 516–571 (2015) 12. Chauhan, A., Vaish, R.: Magnetic material selection using multiple attribute decision making approach. Mater. Des. 36, 1–5 (2012) 13. Zavadskas, E., Skibniewski, M., Antucheviciene, J.: Performance analysis of civil engineering journals based on the web of science database. Arch. Civil Mech. Eng. 14, 519–527 (2014) 14. Pongprasert, P., Kubota, H.: TOD residents’ attitudes toward walking to transit station: a case study of transit-oriented developments (TODs) in Bangkok. Thailand. J. Mod. Transp. 27, 39–51 (2019)

Developing a Transaction System in Blockchain Afsana Nur Meem1 , Lamya Ishrat Nodi1 , Efte Kharul Islam1 , Minhazul Amin Tomal1 , Ahmed Wasif Reza1(B) , and Mohammad Shamsul Arefin2,3(B) 1 Department of CSE, East West University, Dhaka, Bangladesh

[email protected]

2 Department of CSE, Daffodil International University, Dhaka, Bangladesh

[email protected] 3 Department of CSE, Chittagong University of Engineering and Technology, Chattogram,

Bangladesh

Abstract. Blockchain technology is advancing quickly and does not appear to be slowing down. Many things that looked inconceivable in the last few decades— like excessive transaction costs, double spending, net fraud, recovering lost data, etc.—turned out to be true. But due to blockchain technology, all of this may now be avoided. Every block on the blockchain includes some information and the cryptographic hash of the previous block. This system is reliable. At present, various types of transactions are done using various types of systems. This system can be replaced by such. Current online transaction gateways are vulnerable to hacking, allowing attackers to manipulate the network and cause financial loss. As transactions have to pass through multiple transaction systems, which takes time and carries the risk of transaction failure. Based on our research, it is possible to conclude that blockchain is the best technology for cryptocurrencies in commercial transactions, as it enables cryptocurrencies to operate decentralized. This can reduce risk as well as transaction costs. So, our system would change the security change the current banking system is some criticism that there are uses that are not legal and also some community affects the research circle has curiosity in it. Keywords: Blockchain · Transaction system · Cryptocurrency · Ethereum

1 Introduction To create the history of any digital asset or currency unchangeable and accessible “Blockchain” is a well-known technology which is also called DLT (Distributed Ledger Technology) where cryptographic hashing takes place. It is much faster, better, and more precise. This technology delivers instantaneous, shareable, and completely transparent data that is recorded on blockchain networks and can only be viewed by the network users who have permission for that, we can say that the blockchain is great for providing such information [1]. Accounts, orders, payments, productions, and among other things can be handled by the blockchain network. Moreover, because all members have a single understanding of the truth, you may view a transaction from beginning to end, giving you a firm conviction as well as additional benefits and opportunities. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 262–276, 2024. https://doi.org/10.1007/978-3-031-50158-6_27

Developing a Transaction System in Blockchain

263

Blockchain relies on cryptography to secure all data and value storage and ensure that transactions are conducted safely. Many different academic fields have investigated blockchain technology. For example, a number of academics have looked at the underlying technologies of the blockchain, such as distributed storage, peer-to-peer networking, cryptography, smart contracts, and consensus methods. Several academics, for example, have looked at the underlying technologies, such as cloud databases, peer-to-peer connectivity, cryptography, crypto algorithms, and mutual authentication. Bitcoin is a means of trade, a reserve of value, and a measurement unit of cryptocurrency. Bitcoin was originally developed using a blockchain. It was not until 2014 that people understood the benefits of blockchain, which extends beyond cryptocurrencies [2]. Bitcoin is a digital currency (a form of payment), but it can also be considered a speculation commodity (how much is it valued). It was created in 2009 and is generally acknowledged as the first digital asset. Crypto assets, or digital assets, are digital representations of value accessible by encryption and blockchain. Ethereum may be a blockchain-based computing platform that enables developers to develop and deploy autonomous applications. Systems that are not being run by one authority will be able to produce a distributed application during which the participants square measure the decision-making authority. The awareness of blockchain and cryptocurrencies has recently expanded exponentially, though in various directions according to the viewpoints of the users. Several people became users and investors in cryptocurrencies such as Bitcoin and Ethereum, which have opened the way for the institution of tiny and medium companies [3]. The drawbacks of current systems are addressed by the blockchain transaction system we offer in this article, which is built on Python. Despite the fact that there have been a number of studies in this field in recent years, none of them, as far as we are aware, have used Python to construct a blockchain transaction system. Our technology fills this knowledge gap and offers a practical and effective way to execute blockchain transactions in Python. Moreover, we aim to build a transaction system in blockchain that follows the powerefficient algorithms of blockchain which is proof of stake. And to build a transaction system to transact among the users of the system with the help of a bank (which is the source of cryptocurrency in our system). To reach our goal, we need to be specific with some actions like building a system that’s going to be more power efficient, we need to follow the proof of stake algorithm. And if we speak in detail then let’s see some points below there: 1. We first need to define some modules to create the system in Python. 2. We must define our database as we cannot develop this through the local database. The flask could be the best option for this. 3. In the code, we will create the log-in, register, and log-out options (just to make it look smarter). 4. We will create the first block after that, and to get the valid transaction, we must input a line of codes to test if the transaction hash of the previous user can be the new hash of the present user. 5. We will define the blockchain on our app page.

264

A. N. Meem et al.

6. We will create a method on the app page that will make the transaction available for different users and also create a buy page so that a user can buy Ethereum from the bank through our system. 7. And all procedures will be done under the proof-of-stake algorithm, to make our system an example of Green IT. The key contributions of this paper are as follows: 1. Accomplished the limitation of existing paper which could not implement any blockchain transaction system using Python. 2. Supports green IT as it has been built based on a power-efficient algorithm. 3. By providing valid transactions, our system proved its authenticity in the field of Information Technology. The body of the paper is structured as follows: Sect. 1 introduces this research study, while Sect. 2 provides a brief assessment of related publications on the Blockchain transaction system, Sect. 3 provides a transaction model, flow diagram, system diagram, algorithms, and calculation, Sect. 4 explains about the comparison of related works with the proposed model, implementation and finally, Sect. 5 concludes with the limitations of our research.

2 Related Work This section focuses on existing and related blockchain transaction work. The authors have spent most of their previous publications researching how to create a transaction mechanism. In [4] the two most pioneering Ethereum clients are Geth and Parity. In this study, we decided to explore Ethereum’s performance under these two different clients. Geth is one of the first Ethereum clients (refer to Table 1) to be implemented in the Go programming language. Geth is easy to build, and it has a mining option on its own. Parity is an Ethereum newer client with an emphasis on efficiency. Recently, this client has attracted the attention of many Ethereum decentralized developers due to the faster sync process. This document [5] goes into depth about the real-world use scenario, the idea, the final implementation, and the outcomes. It intends to serve as a guide for others by highlighting potential challenges and opportunities when implementing a blockchain in a sector other than financial transactions. This paper does not develop the blockchain system or the desired transaction mechanism. This paper has guidelines, it has the previous implementation records and some very nice designs but does not give us the system and we know that for more efficiency we need a system. We can assume and calculate many things by using the system, but with only a design of a transaction mechanism, it is quite hard to think of how the system will work. A transaction system is a sensitive one because there is a matter of value, and we know that money is valuable and important to us. A secured system is all we need to build a proper transaction system and without security, the system will be a clueless one. Additionally, this paper used to scratch, whereas our system and paper used Python, which is considerably simpler and intelligible and meets this paper’s limitations.

Developing a Transaction System in Blockchain

265

Table 1. Ethereum clients Client

Language

Developers

Go-Ethereum (Geth)

Go

Ethereum foundation

Parity

Rust

Ethcore

Cpp-Ethereum

C++

Ethereum foundation

Ethereumjs-lib

Javascript

Ethereum foundation

Ethereum

Java

ether.camp

Ruby-Ethereum

Ruby

Jan Xie

EthereumH

Haskell

BlockApp

In [6], they begin with a review of blockchain technology, including fundamental ideas, operations, advantages, and applications. Consensual methods are then described briefly, and the Proof-of-Work (PoW) mechanism and its existing difficulties are explained. Following that, they provide an important emergent Proof-of-Stake (PoS) consensus block. This paper has some serious comparisons between proof-of-stake and proof-of-work algorithms. These are two different types of blockchain algorithms that are used in different types of cryptocurrencies. By using the proof of work algorithm huge amounts of energy are being wasted all over the world and less amount of power is being wasted in the proof of stake algorithm. And it explains how the blockchain’s algorithms work. After reading the paper we may infer that it encourages us to work on the blockchain’s effective algorithm. Utilizing their research, one may build a system by choosing which algorithm they will work on. But there is still no implementation of the system that is important for any kind of algorithm and design. And we would also love to say that our papers also complete the limitation of this paper. In numerous areas, our approach is more effective than the one outlined in the reference [7]. First off, we used Python, a programming language renowned for its effectiveness and user-friendliness. Second, our method calculates the transaction data more efficiently, leading to quicker processing times. Last but not least, we thoroughly evaluated our system using a variety of transaction instances to make sure it is both trustworthy and efficient. The limitations of these papers and articles are that they provide us with a detailed model for developing a blockchain or transaction system and also provide us with some comparison between proof of stake and proof of work, but they do not provide any transaction systems that we have overcome in our paper. Again, we have seen that the transaction system algorithm and system are absent. By observing a system, we can understand how the money will be transacted and also how secure the system is. We have a proper transaction system in the blockchain that we have developed. The name of our system is Ethereum (ETH) Cryptocurrency.

266

A. N. Meem et al.

3 Materials and Methods To build a transaction system on the blockchain, we have chosen to follow the proof-ofstake algorithm over proof-of-work, as we know that it is much more eco-friendly and less power-consuming. For materials, we have used. • Python and JavaScript for implementing the whole transaction system. We tried to keep it look more convenient. That is why we did not go for scratch or solidity. • CSS and HTML for designing the system and the page. We have used SCSS particularly. • The Django framework we have used to complete this system makes it so much easier to use and understand at the same time. Proof-of-stake (PoS) can be a more energy-efficient replacement for PoW. The miner does not need to spend a huge amount of computational power to solve the theoretical challenge using this consensus methodology. Instead, participation in the block generation stage is required to have a sufficient interest in the system.

Fig. 1. The transaction model for blockchain.

Consider Fig. 1: You decide to pay your friend George in Ethereum for a pizza that you ordered from him. You produce and publish an entry on the Ethereum blockchain when you send George money in Ethereum. The network’s other computers will verify that you have not already sent the information that represents Ethereum to another user, stopping you from using a digital currency that has already been used. Each computer in the Ethereum network maintains a log of all network transactions and keeps tabs on each account’s balance [8]. For measuring truth and allowing both consumers and producers to certify that their data are accurate and have not been modified blockchain offers the framework for it. For example, if a blockchain has 10 blocks, a block numbered 10 provides the hash

Developing a Transaction System in Blockchain

267

of the previous subsequent block, and the contents of the current block are utilized to create a brand-new block [9]. As a result, all of the blocks in the existing chain are linked and connected. Even the transactions are connected. Now, a simple modification of any transaction will drastically alter the hash [10]. If someone intends to modify any information, he must change all of the prior block’s hash data, which is considered an enormously difficult undertaking given the amount of work that needs to be done. Furthermore, once a miner produces a block, it is acknowledged by other network users [11]. As a result, any type of modification or manipulation will be identified by the network. As a result, the blockchain is extremely tamperproof and is considered an immutable distributed network.

Fig. 2. Transaction flow diagram.

In Fig. 2, we are trying to show the exact flow of how a block or coin is being made and transferred to our user’s account. First, the header of the block is generated by giving its required information. That is the time of mining. Exactly after that, a cache using sha256 computes a 16 MB operation and passes it to the next step to store the dataset it collects storage. All this process continues according to the sha256 algorithm [12]. On the other hand, none gets selected and as per the algorithm it increments and transforms into random slices. When the threshold gets bigger than the result it generates a new block.

268

A. N. Meem et al.

A transactional flow is a form of data flow diagram that illustrates the process of transactions within an organization or department. Transactional flow diagrams are used to provide an overview of the process and allow for further expansion.

Fig. 3. System diagram.

In Fig. 3, we are trying to show the exact flow of our system. When a user logs in to our system by providing valid credentials, he/she directly transfers into the dashboard/homepage. But if he/she fails to provide the required info, he/she directly returns to the register/signup page. And after a successful log-in, he/she can go to any page consisting of the transaction page and buy page just by logging in once. Users can transact or buy just by being in the system with sufficient amounts. And the user can log out of the system on any page of the system.

Developing a Transaction System in Blockchain

269

A system diagram is a basic and exceptionally high description of an existing or tobe-built system. It is a basic diagram that can be created interactively in a short period. It can support a team in obtaining clear, accurate, and shared knowledge of a system. To implement our proposed model, these below steps are needed. First the procedures for implementing our entire model: Procedure: 1: 2: 3: 4: 5: 6: 7: 8:

Log in Create an account Transact with another user Buy coins Provide valid credentials Creating hash Creating a unique block Add and delete blocks

To solve our proposed method and automate the solution, we have included the following algorithms. Algorithm 1: wrap to define if the user is currently logged in from session // called when a user tries to log in Inputs: 1: login credentials given () 2: session.check password() 3: if (credentials == right) { 4: show homepage(); 5: } else { 6: return to sign up page;} Output: Authenticity of the user is examined in this section

Algorithm 2: Registration page function // called when a new user tries to create an account Inputs: 1: user info is given () 2: session.check info() 3: if (info == valid) 4: user created (login page) a. Check the validity of the user information b. If the information is valid, create a new user account and direct the user to the login page c. If the information is invalid, prompt the user to input valid information 5: else: 6: input info as per command (register page) Output: defining registration/sign-up page

270

A. N. Meem et al.

Algorithm 3: Transaction page // called when a user tries to transact with another user of this very same system Inputs: 1: request transaction 2: session.get user_info 3: check_balance 4: if (balance = valid) 5: coins transfer to the user () a. Transfer coins to the other user b. Update the transaction history 6: else 7: not enough/invalid amount 8: return to transaction page () Output: The transaction successfully executes in this section

Algorithm 4: Buy page // called when a user tries to buy coins from the bank through our system Inputs: 1: request buy (Ethereum coin) 2: session.get user info 3: check_validity of user 4: if (user = valid && amount == sufficient) a. Add coins to user’s account b. Display success message 5: coins added to users account () 6: else 7: insufficient balance/invalid info 8: return to homepage () Output: Successfully coin purchased from the bank in this section

Algorithm 5: Home page // called when a user logged in providing valid credentials Inputs: 1: session.log_in(user_id, password) 2: user_info = session.check_user_info() 3: if user_info[‘coin_balance’] > 0: 4: display_coin_balance(user_info[‘coin_balance’]) 5: else: 6: display_username(user_info[‘username’]) Output: dashboard or homepage is shown to the user right after providing valid credentials

Algorithm 6: creating hash // called up initially when a user tries to buy or transact coins (continued)

Developing a Transaction System in Blockchain

271

(continued) Inputs: 1: request transaction 2: creating block(sha256) 3: checking_activity 3: if (previous hash == found) 5: create new hash () 4: else 5: prefer creating unique block () Outputs: Hash creation occurs when a user tries to buy or transact a coin by looping each argument

Algorithm 7: creating unique block // called when a different user tries to transact Inputs: 1: request transaction 2: creating block(sha256) 3: checking_activity 4: self.previous_hash = previous_hash 5: self.nonce = nonce 6: updating hash a. Set self.previous_hash to the previous hash value b. Set self.nonce to a randomly generated number and update the hash 7: return blockchain () Output: unique block creates and hash updates in this session

Algorithm 8: Add/delete blocks Inputs: 1: request transaction(bank) 2: checking_activity 3: if (previous block == found || unique block = prepared 4: new block added 3: else 4: delete the corrupted block () Output: To get the valid transaction this section makes your new block and delete for the same purpose as well

Ethereum employs a particular elliptic curve and a set of theoretical constants described in the secp256k1 standards specified by the US National Institute of Standards and Technology (NIST). The secp256k1 curve is defined by the elliptic curve produced by the following function: y2 = (x3 + 7) over (F p)

(1)

y2 mod p = (x3 + 7) mod p

(2)

or:

The mod p (modulo prime number p) which indicates that the above curve is over a finite field of prime order p, also written as F p, where p = 2256 – 232 – 29 – 28 – 27 – 26 – 24 – 1, that means this is a very huge prime number.

272

A. N. Meem et al.

Fig. 4. Visualization of an elliptic curve.

Over a finite field of prime order rather than over the real numbers the curve is defined, it looks like a pattern of dots scattered in two dimensions and for this reason, it is quite difficult to visualize. However, math is identical to that of an elliptical curve with real numbers. As an example, Fig. 4 shows the same elliptic curve over a much smaller finite field of prime order 17, showing a pattern of dots on a grid. The secp256k1 Ethereum elliptic curve can be thought of as a much more complex pattern of dots on an unfathomably large grid. It shows us our transaction method. How to block is being made and the transaction is happening.

4 Result and Discussion To compare the description, aims, and limitations, we have prepared a comparison table (Table 2) to classify the categories. Here, we have discussed the descriptions, aims, and limitations of related papers and compared them with our paper. From the above table, we have seen that no paper has done a transaction system where we build a proper system. However, some of the papers do not have proper calculations and systems which we have overcome in our paper. Our main goal was to implement a proper transaction system with no limitations. For the system, we focus on implementation which will support Green IT and minimize transaction risk. An activity started by an externally owned account, or an account maintained by a person rather than a contract is referred to as an Ethereum transaction. We carried out multiple transactions and compared them to those of existing systems to validate the proposed transaction system. We discovered that our approach effectively and efficiently reduced transaction risks. Our suggested solution also addressed

Developing a Transaction System in Blockchain

273

Table 2. A comparison of related works. References

Concept

Aims

Limitations

[4]

Most of the pioneer clients are Geth and Parity

Faster sync process

Does not develop any system

[5]

Real-world use case

Serving as a guide to others by showing potential

Does not give a proper efficient calculation

[6]

Provide an overview of Utilizing their research blockchain technology

Still no implementation of the system

[7]

Provide a detailed model for the blockchain transaction system

Implement a system

All of the systems will not work as they give

Proposed model

Build a transaction system

Minimizing the markets transaction risk through our system

We have no limitations and also, we have a proper system

the drawbacks listed in Table 2, such as poor implementation and inaccurate computations. Overall, our technology has shown potential for offering a trustworthy and secure platform for transactions. Here, we have shown below a flow diagram that gives a piece of basic information about our Ethereum transaction and how it works.

Fig. 5. Ethereum transaction.

274

A. N. Meem et al.

Figure 5 shows us our transaction method. How the block is being made and what the transaction is happening. From the above flow diagram, we can see that 100 Ethereum is being transferred to the user from the bank. And (000011a1410e75a409b404c0f5081dcfac9e15b8d77b1ba5c3d9f46303c7aca4) is transaction ID. Ethereum defines itself as “the world’s programmatic blockchain”, offering its services as an electronic, programmable network on which anybody may construct coins and autonomous applications [13]. Unlike Bitcoin, which has a maximum circulation of 21 million coins, the amount of ETH that can be produced is unlimited, although the time it takes to process each ETH block restricts the number of ethers that can be generated each year [14]. Another contrast between Ethereum and Bitcoin is how the network handles transaction management fees. Contemporary technology, as are modern issues [15]. A significant amount of electricity is required to use Blockchain technology. No one is taking measures to prevent this problem and make it more energy efficient due to new technologies [16]. So, to overcome this problem, we have built an algorithm and also a system where people can transact their cryptocurrency using our system. It is not only a secure system for transactions but also an energy-efficient system where a huge amount of power can be saved, and it will be much more beneficial for our modern environment. Our algorithm is being developed using the Python programming language instead of a blockchain. Questions can arise that the system is not very secure, but we have proven that by using a programming language like Python, we can save an enormous amount of power and also can do our daily cryptocurrency transactions.

5 Disadvantages of Using Blockchain While blockchain technology has many advantages, it also has certain disadvantages. Scalability of the blockchain is one of the main issues. The amount of computational power needed to validate transactions grows together with the network’s node count. It can result in longer transaction times and greater costs, which makes it less useful for usage in some applications. The energy use of blockchain is another problem. The processing power is very high if we want to mine new coins, which are then used to validate the transactions and then it will add some new blocks into the chain. This has raised questions about how blockchain may affect the environment, especially in light of the rising need for energy-efficient products.

6 Costs of Using Blockchain There are a number of expenses related to blockchain technology. First off, because the infrastructure takes a lot of computer power and energy, the initial cost of building it up might be substantial. However, especially for large-scale deployments, the cost of maintaining and upgrading the system might be significant. A further expense is the transaction fee, which is paid to the network nodes in order for the transaction to be verified and recorded on the blockchain. Depending on network congestion, this charge can change and can be costly, especially for high-value transactions.

Developing a Transaction System in Blockchain

275

7 Conclusion Our sole purpose of this paper is to change the current banking system securely. The system enables transactions that will be done online between two people of parties and it will be based on cryptographic proof and also the transaction will not rely on and trust any third party. In the present world, energy is being wasted enormously for various kinds of reasons. We can set the power consumption using blockchain at the top of the list. For this, power is wasted more than a whole country like Argentina annually. To make our world more advanced we cannot stop the use of blockchain but replace it with a more energy-efficient and secure system. Our system is fully matched with that kind of system which we need to overcome the power consumption. Nowadays, people transact their money all over the world using cryptocurrency. Behind all of the crypto, there is one algorithm called the blockchain. Using this, the currencies are transacted in a very secure way. Sooner all the banks will move on to cryptocurrency too. But there is a huge problem with it, which is that it uses a huge amount of power, which is not good from the perspective of green IT. So, keeping these points in mind, we have created a transaction system with lower power consumption. Our suggested approach makes use of green IT concepts to cut down on energy use and environmental effect. We have created a system that is not only efficient at managing blockchain transactions but also sustainable by adopting effective algorithms and using Python as a programming language. Moreover, we have addressed the shortcomings of current models by offering a thorough blockchain data collecting and administration system that may successfully reduce transaction risks. The development of a reliable and effective blockchain transaction system that may lessen the environmental impact of this developing technology is advanced significantly by our suggested approach.

References 1. Bassey, E.: Elice Bassey, 1 Sept 2022. Retrieved 12 Sept 2022 2. Kelley, K.: What is Ethereum? Explained with Features and Applications. Simplilearn.com, 20 July 2022. Retrieved 12 Sept 2022 3. Bentov, I., Lee, C., Mizrahi, A., Rosenfeld, M.: Proof of activity: extending bitcoin’s proof of work via proof of stake [extended abstract]y. ACM SIGMETRICS Perform. Eval. Rev. 42(3), 34–37 (2014) 4. Rouhani, S.: Ethereum Transaction In Private Blockchain, 04 Feb 2021 5. Knirsch, F.: Implementing a blockchain from scratch: why, how, and what we learned. EURASIP J. Inform. Secur. (2019). Retrieved 12 Sept 2022 6. Cong, T., Dinh, T.: Proof-of-stake consensus mechanisms for future blockchain networks: fundamentals, applications and opportunities. IEEE Access 7, 112411–112434 (2019). https:// doi.org/10.1109/ACCESS.2019.2933326 7. Hu, W.: A blockchain-based secure transaction model for distributed energy in Industrial Internet of things. Alex. Eng. J. 60(1):491–500 (2021) 8. Erratum. Acad. Emerg. Med. 26(4), 462–462 (2019) 9. Shrier, D., Wu, W., Pentland, A.: Blockchain and Infrastructure (Identity Data Security), vol 1(3). Cambridge, MA, USA (2016)

276

A. N. Meem et al.

10. Karandikar, N., Chakravorty, A., Rong, C.: Blockchain based transaction system with fungible and non-fungible tokens for a community-based energy infrastructure. Sensors 21(11), 3822 (2021) 11. Feng, Q., He, D., Zeadally, S., Khan, M.K., Kumar, N.: A survey on privacy protection in blockchain system. J. Netw. Comput. Appl. 126, 45–58 (2019) 12. Peter, H., Moser, A.: Blockchain-applications in banking & payment transactions: Results of a survey. Eur. Financ. Syst. 141, 141 (2017) 13. Cocco, L., Pinna, A., Marchesi, M.: Banking on blockchain: costs savings thanks to the blockchain technology. Future Int. 9(3), 25 (2017) 14. Tasatanattakool, P.: Blockchain: challenges and applications. In: IEEE Conference Publication. IEEE Xplore. Retrieved 17 Sept 2022 15. Treleaven, P., Brown, R.G., Yang, D.: Blockchain technology in finance. Computer 50(9), 14–17 (2017) 16. Nguyen, Q.K.: Blockchain—a financial technology for future sustainable development. In: 2016 3rd International Conference on Green Technology and Sustainable Development (GTSD), pp. 51–54. IEEE (2016)

Using the Phi-Function Technique for the Optimized Virtual Localization Problem Sergiy Plankovskyy1 , Yevgen Tsegelnyk1(B) , Tetyana Romanova2,3 , Oleksandr Pankratov2 , Igor Litvinchev4 , and Volodymyr Kombarov1 1 O. M. Beketov National University of Urban Economy in Kharkiv, 17 Marshala Bazhanova

Street, Kharkiv 61002, Ukraine [email protected] 2 Anatolii Pidhornyi Institute of Mechanical Engineering, Problems of the National Academy of Sciences of Ukraine, 2/10 Pozharskogo Str., Kharkiv 61046, Ukraine 3 University of Leeds, Maurice Keyworth Building, Leeds LS2 9JT, UK 4 Faculty of Mechanical and Electrical Engineering, Nuevo Leon State University, 66450 Monterrey, NL, Mexico

Abstract. Optimized virtual localization arises in manufacturing parts from “nearly shaped” workpieces using computer numerical control (CNC) machining. To reduce the material consumption and duration of machining processes, a part (a polygonal object) must be placed completely in a workpiece (a polygonal domain) maximizing the distance between the part and the boundary of the workpiece. Using the phi-function technique a continuous nonlinear model is constructed for this non-standard packing problem and a corresponding solution strategy is proposed. Numerical examples are provided to demonstrate the efficiency of the proposed approach. Keywords: Polygonal workpiece · Virtual localization · Phi-functions · CNC machining

1 Introduction Technological processes for obtaining finished parts from workpieces with minimal allowances for machining [1, 2], e.g., precision casting, stamping, or additive manufacturing (3D printing) [3, 4], are widely used in practice. The use of such workpieces allows a significant reduction in the consumption of material and the duration of machining operations. Correspondingly, the problem of smart localizing the parts in the workpieces arises. This is especially important for high-weighted parts/workpieces and those having large dimensions or low rigidity [5, 6]. Traditionally, during machining, localization is carried out with the help of special jigs that not only fix the workpiece but also ensure its certain orientation in the machine coordinate system. However, in many cases, especially in the manufacturing of complex shape large-sized parts, the design, and manufacturing of such jigs require more time and money than the part manufacturing, and the localization process requires considerable © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 277–287, 2024. https://doi.org/10.1007/978-3-031-50158-6_28

278

S. Plankovskyy et al.

time and is associated with significant difficulties for heavy parts. An alternative is using adaptive machining [7] with virtual localization. In this case, the workpiece is installed on the machine table with the simplified jigs, and its orientation in the machine coordinate system and the shape is determined using built-in measuring systems. Thus, the idea of adaptive machining with virtual localization is that instead of fixing the workpiece in a theoretically specified position, a computer numerical control (CNC) correction program is used for its current position. This approach is successfully used in the aerospace industry for the machining of large-sized thin-walled parts [8–10], the manufacturing and restoration of gas turbine engine blades [11–13], the nozzles of liquid rocket engines [14, 15], as well as in the manufacturing of complex shape large-sized parts, such as paddles propellers, hydraulic and steam turbine blades, etc. The solution of the localization problem in almost all studies devoted to the issues of virtual localization is conducted in two stages. At the first stage, preliminary localization of the CAD model inside the workpiece is carried out. The starting position of the CAD model in the majority of papers is determined by the condition of the alignment of the centers of gravity of the thin shells that coincide with the surfaces of the workpiece and the CAD model of the part [16–18]. The starting localization option proposed in [19] is more accurate in the case when the machining allowance is approximately the same over the entire surface of the part. In this case, it is suggested to combine not only the centers of gravity of the shells, but also their main central inertia axes. At the second stage of solving the virtual localization problem, accurate location is performed by translating and rotating the CAD model relative to its initial position. For this, in the vast majority of papers, the iterative Iterative Closest Point (ICP) algorithm proposed in [20] is used. The criterion for the optimal mutual placement of geometric objects in this method is the minimization of the sum of squared distances between their elements (for the problem of virtual localization – distances between points from the scanned cloud and curves and surfaces of the CAD model). Some modifications of the ICP algorithm use global control points calculated as average values according to coordinates [21], or special points obtained on the basis of Bearing Angle Images [22], the use of which is not requires setting initial approximations for the rotation matrix and the displacement vector, thereby allowing to avoid hitting a local minimum, requires fewer iterations for the alignment of point clouds, and is performed faster compared to the basic algorithm. Besides ICP algorithms, several other methods have been proposed. In [23], a method based on comparisons of obtained surfaces with nominal data or for direct comparison of different scanned surfaces using an algorithm based on extended Gaussian curvature and a method of comparing characteristics based on aggregate normal orientation was studied. In paper [24], the problem of matching three-dimensional objects was considered from the point of view of a nonlinear polynomial equations system solutions, which can be solved using an algorithm, where Gaussian and average curvature were used as evaluation criteria. Nevertheless, for today, the ICP algorithm is essentially the basic standard for the problems of combining geometric objects. The disadvantage of ICP methods in relation to the virtual localization problem is the application of the criterion of the average value of the square of the distance minimization between the cloud of measurement points and the project surface. With such a criterion, the algorithm

Using the Phi-Function Technique for the Optimized

279

is not sensitive to where exactly the point from the transformed cloud is located in relation to the CAD surface – inside or outside the CAD model. Obviously, the situation cannot be changed by using other norms for determining the distance, insensitive to the mutual location of points, for example, the Manhattan norm, which was used in some papers devoted to the modification of the ICP algorithm for the purpose of accelerating convergence [25]. This applies to other similar papers, for example, [26, 27], where a modification of the distance function from the point to the surface was introduced and the properties of the complex error of the surface were studied. From the point of view of the virtual localization problem, it may be promising to use the method of Lagrange multipliers used in [28] when considering the problem of localization a free-form surface. In addition, we should highlight the works [29, 30] in which it is proposed to minimize the functional U = max[ρi ] + UP , where ρi is the distance between the point from the cloud obtained by measurement to the surface of the CAD model; UP is a penalty function. When the coefficients of the penalty function are selected appropriately, it acquires properties that allow to “push” points from the cloud outside the CAD model area. The papers [29, 30] demonstrated the effectiveness of using this approach in the virtual localization problem of complex shape parts. At the same time, in the known papers, a reasonable approach to the formation of penalty functions, which are used when recording the objective function, has not been developed. Thus, the virtual localization problem of complex shape objects on the basis of the combined use of a CAD model and a point cloud obtained during the measurement of the workpiece remains a difficult task. The development of new approaches to solving this problem is expedient to start with the example of two-dimensional problems. In addition to reducing computational costs, this approach can be directly applied to solving virtual localization problems in contour milling [31]. As shown in [31, 32], for computer modeling of the parts contour milling problem on CNC machines, the representation of geometric models in the form of broken lines can be successfully used. For this purpose, the flat geometric contours of the workpiece and the part are specified by two-dimensional digital arrays of points with a given step. The position of the workpiece on the machine table is usually determined with the built-in measuring heads, which are capable of automatically determining the coordinates of the points of the workpiece contour in the machine coordinate system with an error of up to 0.001 mm. In Fig. 1 shown the contours of part D and workpiece B, which are represented by digital arrays containing the coordinates of two sets of points D = {d1 , d2 , . . . , dn } and B = {b1 , b2 , . . . , bm }. The virtual localization problem is aiming at finding placement parameters that ensure the minimization of the maximum machining allowance (that is, the distance between the part and workpiece contours). Note that even in such a simplified setting when applying the algorithm based on the maximization of the Hausdorff distance, in [31] cases of intersection of the contours of the part and the workpiece were reported. To prevent such errors, virtual localization in [31] provided for the final decision to complete the virtual localization procedure by the technological engineer. Thus, the task of virtual localization during CNC machining requires the development of new algorithms that would enable it to be carried out without errors and in automatic

280

S. Plankovskyy et al. Y D

b6

d1 b1 B 0

d2 b2

b5 d3 d4 b3

b4 X

Fig. 1. An example of virtual localization during contour milling: D is the contour of the part, B is the contour of the workpiece [31]

mode. The paper considers an approach that uses the phi-functions technique [33–35] to solve the virtual localization problem.

2 Problem Formulation Let B ⊂ R2 be a fixed bounded covex polygonal domain (workpiece) given by their verticies uk = (xk , yk ) for k = 1, . . . , n, and B = {(x, y) : ϕk (x, y) ≥ 0, k = 1, . . . , n},   where ϕk (x, y) = αk x − βk y + γk , α2k + β2k = 1, αk = (yk+1 − yk )/Ak , βk = (xk+1 −    xk )/Ak , γk = −αk xk − βk yk , Ak = (xk+1 − xk )2 + (yk+1 − yk )2 , for k = 1, . . . , n, subject to un+1 = u1 . With each irregular part we accociate its polygonal convex hull D ⊂ R2 given by their verticies v˜ i = (˜xi , y˜ i ), i = 1, . . . , m. The location and orientation of D is defined by a variable vector of its placement parameters (xd , yd , θ). A translation of D by the vector vd = (xd , yd ) and a rotation of D trough the angle θ ∈ [0, 2π) is defined as

D(vd , θ) = {t ∈ R2 : t = vd + M (θ)˜t , ∀ ˜t ∈ D}, where M (θ) is a standard rotation matrix. Thus each point ˜t = (˜x, y˜ ) ∈ D in the local coordinate system of D is transformed into point t = (x, y), where x = x˜ · cos θ + y˜ · sin θ + xd , y = −˜x · sin θ + y˜ · cos θ + yd . Optimized virtual localization problem. Place a given object D completely inside a fixed polygonal domain (workpiece) B maximizing Euclidean distance between the object (part) D and the boundary of B. Denote the variable Euclidean distance between the object D and the boundary of B by ρ, i.e. ρ = dist(D, B∗ ) = min ∗ d − b, B∗ = R2 /int(B). Therefore, the d ∈D,b∈B

placement problem is aiming to search for a vector (xd , yd , θ) maximizing ρ.

Using the Phi-Function Technique for the Optimized

281

3 Mathematical Model Using the phi-function technique [33–35] the optimized virtual localization problem can be formulated as the following nonlinear programming model: max

(vd ,θ,ρ)∈W

ρ

(1)

subject to  ∗    W = (vd , θ, ρ) ∈ R4 B D (vd , θ) ≥ ρ , ρ ≥ 0 .

(2) ∗

In the model (vd , θ, ρ) = (xd , yd , θ, ρ) is a vector of variables, B D (vd , θ) is the normalized phi-function for objects D(vd , θ) and B∗ defined in the following form: ∗

B D (vd , θ) =

min

k=1,...,n,i=1,...,m

ϕk (xdi , ydi ),

ϕk (xdi , ydi ) = αk xdi − βk ydi + γk , xdi = x˜ di · cos θ + y˜ di · sin θ + xd , ydi = −˜xdi · sin θ + y˜ di · cos θ + yd . The variable ρ can be considered as ρ =

min

i=1,...,m,k=1,...,n

ϕk (xdi , ydi ).

4 Solution Strategy The solution algorithm involves three main stages: Stage 1. Constructing a set of feasible starting points of the problem (1)–(2). Stage 2. Searching for a local-optimal maximum of the problem (1)–(2) using IPOPT for each starting point found at Stage 1. Stage 3. Choosing the best local optimal solution from those found at Stage 2. The heuristic algorithm to search for feasible starting point of the problem (1)–(2) involves the following principal steps: Step 1. Generate point vd0 = (xd0 , yd0 ) ∈ B randomly. Step 2. Generate a rotation angle θ0 ∈ [0, 2π] randomly. Step 3. Solve the following nonlinear programming subproblem starting from the point (xd0 , yd0 , θ0 ): max η

(3)

subject to αk (η(˜xdi · cos θ + y˜ di · sin θ) + xd ) − βk (η(−˜xdi · sin θ + y˜ di · cos θ) + yd ) + γk ≥ 0, (4)

282

S. Plankovskyy et al.

for k = 1, . . . , n, i = 1, . . . , m, 0 ≤ η ≤ 1,

(5)

where (vd , θ, η) is a vector of variables. Step 4. If η∗ < 1 is a solution of the problem (3)–(5) then goto Step 1 otherwise (η∗ = 1) take point (vd∗ , θ∗ ) as a feasible starting point of the problem (1)–(2).

5 Computational Results In this section numerical results for several problems instances solved by the algorithm are presented in Table 1 and Fig. 2. All experiments were performed using AMD FX(tm)-6100, 3.30 GHz computer, C++ programming language and the operating system Windows 7. The open access solver IPOPT [36] was used for local optimization under default options.

Fig. 2. Local optimal placement: (a) – Example 1; (b) – Example 2; (c) – Example 3; (d) – Example 4; (e) – Example 5; ( f ) – Example 6

6 Conclusions The numerical results demonstrate that using the phi-functions technique is promising for solving virtual localization problems. The proposed approach eliminates false solutions characterized by intersections between the workpiece and part surfaces. Such errors during localization without the operator’s participation will inevitably lead to production defects. The algorithm to solve the 2D virtual localization problem for convex polygonal part and workpiece was proposed. For all problem instances the solution time did not exceed 2.5 s, which fully meets the requirements of further industrial use of the proposed virtual localization method. An interesting direction for the future research is extending the proposed method to problems where the CAD model of the part has an arbitrary shape and/or is formed by Boolean operations with simple geometric shapes. Considering 3D virtual localization problems is also planned in near future. In the proposed approach large-scale nonlinear optimization problems must be solved. To simplify these problems various linearizations of the principal model can be used. In particular, grid approximations of the domains involved in the problem formulation permit reducing approximately the nonlinear continuous packing problem to a linear mixed integer optimization [37–39].

Using the Phi-Function Technique for the Optimized

283

Table 1. Numerical results (xd∗ , yd∗ , θ∗ )

Time, s/iteration

{(xdi , ydi ), i = 4.391339 1, . . . , m = 10} = {(144, 438), (138, 465), (132, 486), (120, 492), (99, 489), (78, 480), (72, 465), (75, 450), (96, 435), (117, 432)

(518.804021, 169.748525, 1.709054)

0.031/1

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

{(xdi , ydi ), i = 4.245713 1, . . . , m = 10} = {(144, − 438), (138, − 465), (132, − 486), (120, − 492), (99, − 489), (78, − 480), (72, − 465), (75, − 450), (96, − 435), (117, − 432)}

(522.010811, 217.167852, − 1.803106)

0.031/1

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

{(xdi , ydi ), i = 4.327072 1, . . . , m = 17} = {(217.5, 380), (207.5, 382.5), (195, 385), (180, 385), (170, 382.5), (165, 380), (157.5, 372.5), (155, 367.5), (153.75, 362.5), (152.5, 342.5), (157.5, 332.5), (162.5, 325), (175, 322.5), (190, 325), (200, 330), (210, 347.5), (215, 365)}

(− 93.555847, 590.598979, − 2.311687)

0.078/3

No.

Domain B

Object D

1

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

2

3

ρ∗

(continued)

284

S. Plankovskyy et al. Table 1. (continued)

No.

Domain B

Object D

ρ∗

(xd∗ , yd∗ , θ∗ )

Time, s/iteration

4

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

{(xdi , ydi ), i = 1, . . . , m = 17} = {(217.5, − 380), (207.5, − 382.5), (195, − 385), (180, − 385), (170, − 382.5), (165, − 380), (157.5, − 372.5), (155, − 367.5), (153.75, − 362.5), (152.5, − 342.5), (157.5, − 332.5), (162.5, − 325), (175, − 322.5), (190, − 325), (200, − 330), (210, − 347.5), (215, − 365)}

5.587589

(100.605448, 613.501998, − 0.609559)

0.203/6

5

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

{(xdi , ydi ), i = 1, . . . , m = 10} = {(33.15, 248.82), (66.19, 239.94), (82.08, 222.47), (79.22, 197.17), (71.08, 176.73), (55.85, 166.60), (40.63, 176.73), (30.07, 195.89), (24.11, 221.43), (27.16, 242.08)}

3.74537

(193.096076, 374.228730, − 3.620032)

0.078/3

6

{(xk , yk ), k = 1, . . . , n = 10} = {(48.76, 263.12), (75.70, 252.60), (81.47, 230.65), (78.82, 201.42), (69.55, 178.14), (52.41, 155.70), (29.12, 178.14), (12.77, 200.58), (9.27, 232.35), (27.74, 251.55)}

{(xdi , ydi ), i = 4.448 1, . . . , m = 10} = {(33.15, − 248.82), (66.19, − 239.94), (82.08, − 222.47), (79.22, − 197.17), (71.08, − 176.73), (55.85, − 166.60), (40.63, − 176.73), (30.07, − 195.89), (24.11, − 221.43), (27.16, − 242.08)}

(151.075074, 23.172399, 3.397443)

0.172/4

Using the Phi-Function Technique for the Optimized

285

Acknowledgements. The research is supported by Volkswagen Foundation (grant #97775), Ministry of Education and Science of Ukraine (scientific research project No. 0121U109639), British Academy (grant #100072).

References 1. Marini, D., Cunningham, D., Corney, J.R.: Near net shape manufacturing of metal: a review of approaches and their evolutions. Proc. Inst. Mech. Eng. Part B J. Eng. Manufact. 232(4), 650–669 (2018). https://doi.org/10.1177/0954405417708220 2. Kurin, M.O.: Investigation of the method of hyperbolic flow around the wedge with free cutting. Metallofiz. Noveishie Tekhnol. 40(7), 859–876 (2018). https://doi.org/10.15407/mfint. 40.07.0859 3. Riveiro, A., del Val, J., Comesaña, R., et al.: Laser additive manufacturing processes for near net shape components. In: Gupta, K. (eds.) Near Net Shape Manufacturing Processes. MFMT, pp. 105–141. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-10579-2_5 4. Kritskiy, D., Pohudina, O., Kovalevskyi, M., et al.: Powder mixtures analysis for laser cladding using OpenCV library. In: Nechyporuk, M., et al. (eds.) Integrated Computer Technologies in Mechanical Engineering—2021 (ICTM 2021). LNNS, vol. 367, pp. 924–937 (2022). https:// doi.org/10.1007/978-3-030-94259-5_72 5. Kondratiev, A., Píštˇek, V., Smovziuk, L., et al.: Stress-strain behaviour of reparable composite panel with step-variable thickness. Polymers 13(21), 3830 (2021). https://doi.org/10.3390/pol ym13213830 6. Dobrotvorskiy, S., Kononenko, S., Basova, Y., et al.: Development of optimum thin-walled parts milling parameters calculation technique. In: Ivanov, V., et al. (eds.) Advances in Design, Simulation and Manufacturing IV. LNME, pp. 343–352. Springer, Cham (2021). https://doi. org/10.1007/978-3-030-77719-7_34 7. Kombarov, V., Sorokin, V., Tsegelnyk, Y., et al.: Numerical control of machining parts from aluminum alloys with sticking minimization. Int. J. Mech. Appl. Mech. 9, 209–216 (2021). https://doi.org/10.17683/ijomam/issue9.30 8. Bi, Q., Huang, N., Zhang, S., et al.: Adaptive machining for curved contour on deformed large skin based on on-machine measurement and isometric mapping. Int. J. Mach. Tools Manuf 136, 34–44 (2019). https://doi.org/10.1016/j.ijmachtools.2018.09.001 9. Del Sol, I., Rivero, A., López de Lacalle, L.N., Gamez, A.J.: Thin-wall machining of light alloys: a review of models and industrial approaches. Materials 12(12), 2012 (2019). https:// doi.org/10.3390/ma12122012 10. Plankovskyy, S., Myntiuk, V., Tsegelnyk, Y., et al.: Analytical methods for determining the static and dynamic behavior of thin-walled structures during machining. In: Shkarlet, S., et al. (eds.) Mathematical Modeling and Simulation of Systems (MODS’2020). AISC, vol. 1265, pp. 82–91. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-58124-4_8 11. Pavlenko, D., Dvirnyk, Y., Przysowa, R.: Advanced materials and technologies for compressor blades of small turbofan engines. Aerospace 8(1), 1–16 (2021). https://doi.org/10.3390/aer ospace8010001 12. Wu, D., et al.: Analysis of machining deformation for adaptive CNC machining technology of near-net-shaped jet engine blade. Int. J. Adv. Manufact. Technol. 104(9), 3383–3400 (2019). https://doi.org/10.1007/s00170-019-03898-6 13. Denkena, B., Boess, V., Nespor, D., Floeter, F., Rust, F.: Engine blade regeneration: a literature review on common technologies in terms of machining. Int. J. Adv. Manufact. Technol. 81(5), 917–924 (2015). https://doi.org/10.1007/s00170-015-7256-2

286

S. Plankovskyy et al.

14. Poznyakov, V.D., et al.: Cold cracking resistance of butt joints in high-strength steels with different welding techniques. Strength Mater. 51(6), 843–851 (2020). https://doi.org/10.1007/ s11223-020-00132-7 15. Gradl, P.R.: Rapid fabrication techniques for liquid rocket channel wall nozzles. In: 52nd AIAA/SAE/ASEE Joint Propulsion Conference, Paper No. 2016-4771. AIAA (2016). https:// doi.org/10.2514/6.2016-4771 16. Shen, B., Huang, G., Mak, K., Wang, X.: A best-fitting algorithm for optimal location of large-scale blanks with free-form surfaces. J. Mater. Process. Technol. 139, 310–314 (2003). https://doi.org/10.1016/S0924-0136(03)00241-3 17. Sun, Y., Wang, X., Guo, D., Liu, J.: Machining localization and quality evaluation of parts with sculptured surfaces using SQP method. Int. J. Adv. Manufact. Technol. 42(11), 1131–1139 (2009). https://doi.org/10.1007/s00170-008-1673-4 18. Mehrad, V., Xue, D., Gu, P.: Robust localization to align measured points on the manufactured surface with design surface for freeform surface inspection. Comput. Aided Des. 53, 90–103 (2014). https://doi.org/10.1016/j.cad.2014.04.003 19. Plankovskyy, S., Nikolaev, A., Shypul, O., et al.: Balance layout problem with the optimized distances between objects. In: Vasant, P., et al. (eds.) Data Analysis and Optimization for Engineering and Computing Problems. EAISICC, pp. 85–93. Springer, Cham (2020). https:// doi.org/10.1007/978-3-030-48149-0_7 20. Besl, P.J., McKay, N.D.: A method for registration of 3-D shapes. IEEE Trans. Pattern Anal. Mach. Intell. 14(2), 239–256 (1992). https://doi.org/10.1109/34.121791 21. Du, S., Xu, Y., Wan, T., et al.: Robust iterative closest point algorithm based on global reference point for rotation invariant registration. PLoS ONE 12(11), e0188039 (2017). https://doi.org/ 10.1371/journal.pone.0188039 22. Lin, C.-C., Tai, Y.-C., Lee, J.-J., Chen, Y.-S.: A novel point cloud registration using 2D image features. EURASIP J. Adv. Sig. Process. 2017(1), 1–11 (2017). https://doi.org/10.1186/s13 634-016-0435-y 23. Orazi, L., Tani, G.: Geometrical inspection of designed and acquired surfaces. Int. J. Adv. Manufact. Technol. 34(1), 149–155 (2007). https://doi.org/10.1007/s00170-006-0587-2 24. Ko, K.H., Maekawa, T., Patrikalakis, N.M.: An algorithm for optimal free-form object matching. Comput. Aided Des. 35(10), 913–923 (2003). https://doi.org/10.1016/S0010-448 5(02)00205-1 25. Mora, H., Mora-Pascual, J.M., Garcia-Garcia, A., Martinez-Gonzalez, P.: Computational analysis of distance operators for the iterative closest point algorithm. PLoS ONE 11(10), e0164694 (2016). https://doi.org/10.1371/journal.pone.0164694 26. Zhu, L., Xiong, Z., Ding, H., Xiong, Y.: A distance function based approach for localization and profile error evaluation of complex surface. J. Manuf. Sci. Eng. 126(3), 542–554 (2004). https://doi.org/10.1115/1.1763186 27. Zhu, L., Zhang, X., Ding, H., Xiong, Y.: Geometry of signed point-to-surface distance function and its application to surface approximation. J. Comput. Inf. Sci. Eng. 10(4), 041003 (2010). https://doi.org/10.1115/1.3510588 28. Sun, Y., Xu, J., Guo, D., Jia, Z.: A unified localization approach for machining allowance optimization of complex curved surfaces. Precis. Eng. 33(4), 516–523 (2009). https://doi.org/ 10.1016/j.precisioneng.2009.02.003 29. Chatelain, J.F., Fortin, C.: A balancing technique for optimal blank part machining. Precis. Eng. 25(1), 13–23 (2001). https://doi.org/10.1016/S0141-6359(00)00050-7 30. Chatelain, J.F.: A level-based optimization algorithm for complex part localization. Precis. Eng. 29(2), 197–207 (2005). https://doi.org/10.1016/j.precisioneng.2004.07.002 31. Petrakov, Y., Shuplietsov, D.: Contour milling programming technology for virtual basing on a CNC machine. Eastern Eur. J. Enterp. Technol. 2(1), 54–60 (2019). https://doi.org/10. 15587/1729-4061.2019.162673

Using the Phi-Function Technique for the Optimized

287

32. Petrakov, Y., Shuplietsov, D.: Programming of adaptive machining for end milling. Mech. Adv. Technol. 1, 34–40 (2017). https://doi.org/10.20535/2521-1943.2017.79.97342 33. Stoyan, Y., Pankratov, A., Romanova, T.: Placement problems for irregular objects: mathematical modeling, optimization and applications. In: Butenko, S., et al. (eds.) Optimization Methods and Applications. SOIA, vol. 130, pp. 521–559. Springer, Cham (2017). https://doi. org/10.1007/978-3-319-68640-0_25 34. Pankratov, A.V., Romanova, T.E., Chugay, A.M.: Optimal packing of convex polytopes using quasi-phi-functions. J. Mech. Eng. 18(2), 55–65 (2015) 35. Romanova, T., Pankratov, A., Litvinchev, I., et al.: Sparsest packing of two-dimensional objects. Int. J. Prod. Res. 59(13), 3900–3915 (2021). https://doi.org/10.1080/00207543.2020. 1755471 36. Wächter, A., Biegler, L.T.: On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming. Math. Program. 106(1), 25–57 (2006). https:// doi.org/10.1007/s10107-004-0559-y 37. Litvinchev, I., Infante, L., Ozuna Espinosa, E.L.: Approximate circle packing in a rectangular container: integer programming formulations and valid inequalities. In: González-Ramírez, R.G., et al. (eds.) Computational Logistics. LNTCS, vol. 8760, pp. 47–60. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-11421-7_4 38. Litvinchev, I., Ozuna Espinosa, E.L.: Integer programming formulations for approximate packing circles in a rectangular container. Math. Probl. Eng. 2014, 317697 (2014). https:// doi.org/10.1155/2014/317697 39. Litvinchev, I., Ozuna, L.: Approximate packing circles in a rectangular container: valid inequalities and nesting. J. Appl. Res. Technol. 12(4), 716–723 (2014). https://doi.org/10. 1016/S1665-6423(14)70088-4

COVID-19 Detection from Chest X-Ray Images Using CNN Models and Deep Learning Nafisha Binte Moin1 , Shamima Sultana1 , Abdullah Al Munem1 , Omar Tawhid Imam2 , Ahmed Wasif Reza1 , and Mohammad Shamsul Arefin3,4(B) 1 Department of Computer Science and Engineering, East West University, Dhaka, Bangladesh

[email protected]

2 Department of Computer Science and Engineering, Bangladesh University of Engineering and

Technology, Dhaka, Bangladesh 3 Department of Computer Science and Engineering, Daffodil International University,

Dhaka 1341, Bangladesh [email protected] 4 Department of Computer Science and Engineering, Chittagong University of Engineering and Technology, Chattogram 4349, Bangladesh

Abstract. The dangerous disease and ailment known as COVID-19 has caused a global pandemic and immeasurable damage to people all around the globe. Chest X-Rays are being used to identify COVID-19 in current times due to their low cost and efficiency. In this paper, we developed Convolutional Neural Network models to detect COVID-19 from Chest X-Ray images. CNN models from the Keras library such as VGG16, VGG19, Xception, ResNet101, ResNet152, Inception, InceptionResNet, MobileNetV2, DenseNet201, NASNetLarge, and EfficientNetB3 have been used to perform experimentations. The CNN models used ImageNet as its pre-trained weights for transfer learning. Additionally, a multilayered self-designed model has been implemented as well to see the performance. A comparative analysis has been completed in order to find the best-performing CNN model for COVID-19 detection in Chest X-Ray images. From the experiments, we found that the proposed CNN gave the best results. Additionally, it has been observed that MobileNetV2, Inception, ResNet101, and VGG16 give the highest accuracy over 99%, while the lowest accuracy is found by EfficientNetB3 at only around 50%. The self-designed multi-layered model gives a training accuracy of 97.22% and a validation accuracy of 96.42%. A significant increase in accuracy and excellent performance has been seen from the CNN models and the proposed framework. Keywords: Convolutional neural network · COVID-19 · Chest X-ray

1 Introduction In current times, Coronavirus and “COVID-19” is a well-known and feared terms all across the globe. Currently, it is known as the most dangerous and widespread disease all around the world causing the deaths of millions and harm to all sectors and part of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 288–300, 2024. https://doi.org/10.1007/978-3-031-50158-6_29

COVID-19 Detection from Chest X-Ray Images

289

regular life. In general definition, Coronavirus disease (shortly COVID-19) is an airborne and infectious viral disease caused by a virus known as the SARS-CoV-2 virus which mainly affects the lungs of the human respiratory system [1]. The symptoms that are more commonly observed for this disease include fever, cough, loss of taste and smell, and tiredness. There may be pains and difficulty breathing, shortness of breath, and chest pain as well. There are other additional symptoms of this disease as well. This dangerous disease has damaged the medical and financial stability of almost all countries in the world and has caused a global pandemic. Over the last two years, many methods and treatments have been discovered and researched to treat and cure this deadly disease. Along with treatment and cure, many steps have also been taken to ensure the prevention of this deadly and epidemic disease. One method for the identification and detection of this viral disease is using XRay images of the Chest or Lungs to detect the virus. This process can speed up the identification process compared to time-consuming testing in clinics [2]. Over the last few years, the field of artificial intelligence has rapidly advanced in the use of image classification. In the case of medical science, a popular image classification Deep Learning model known as the Convolutional Neural Networks (or CNNs) is widely used. CNNs can be used for the detection of diseases in different organs of human beings and even plants [3]. CNN is an effective tool in the field of medical science in tasks such as image classification, segmentation, and localization with its performance outperforming humans for diseases related to the brain, breast, lungs, and other organs [4].

2 Related Work COVID-19 has become a widely known and feared term all across the globe since the year 2020. In order to tackle the issues rising due to COVID-19 many previous works have been done in recent years to work on fast prediction and detection of the disease. In most earlier related works, the majority of authors worked on improving the performance of the CNNs for the diagnosis of the disease. Panwar et al. [5] used an open-source dataset to conduct experimentations on the proposed model ‘nCOVnet’ with transfer learning. The designed model used VGG16 as its base model and ImageNet as the pre-trained weights. Training accuracy gives 97% accuracy with 98.68% confidence. Jahid et al. [6] performed experimentations on COVID-19 X-Rays, Pneumonia (another lung disease) X-Rays, and unaffected Chest X-Rays using the models three CNNs. The dataset was from Kaggle, where the images were resized with augmentations. The models’ performances were measured by calculating the precision, recall, and other scores of the models which were between the ranges of 0.98–1. Basu et al. [7], it can be seen that CNNs have been used for the identification of COVID-19 images collected from four image databases. Experimentations on the CNN models such as AlexNet (82.98% accuracy), VGGNet (90.13% accuracy), and Resnet (85.98% accuracy) were performed. Ismael et al. [8] work on fine-tuning along with end-to-end training of CNNs for disease detection. ResNet and VGG were the two CNN models which were used where

290

N. B. Moin et al.

feature extraction was performed by using different kernels of SVM. The accuracies of the models range from 85.26 to 92.63%. Hussain et al. [9] proposed a model (with 22 layers) using X-Rays and CTs of Normal, COVID-19, and Pneumonia as input. CoroDet model gives high accuracy for both training and validation along with high confidence scores. Nayak et al. [10] perform experiments to test the performance of CNNs for the identification of diseased lungs. The CNNs used namely, AlexNet, VGG16, GoogleNet, MobileNetV2, SqeezeNet, ResNet34, ResNet50, and InceptionV3 has been used to perform the experiments. The paper shows the model ResNet34 performs better than the rest of the models with an accuracy of 98.33%. Heidari et al. [11] showed that work has been done to improve CNN predictions. Image preprocessing techniques such as histogram equalization algorithms and bilateral low-pass filters are used. The dataset images are used to form pseudo-colored images. The study yields a high confidence interval and high sensitivity. Minaee et al. [12] used transfer learning techniques. Using Chest radiology images from open datasets experiments were performed on the CNN models namely ResNet (18 and 50), as well as SqueezeNet, and also DenseNet-121. The models received a sensitivity of 98% and a specificity rate of 90%. Mangal et al. [13] introduced a COVID detection CNN model called CovidAID. The paper works on the use of X-Rays for further testing of RT-PCR. The model achieves an accuracy of around 90.5% after it has been tested on a publically available dataset. Alazab et al. [14] created datasets that have been used to create the COVID detection model. The models yield an F1 score of around 95–99%. In addition, multiple deep learning methods like the prophet algorithm, ARIMA, and LSTM were implemented to carry out predictions. Wang et al. [15] used deep learning methods. The model was trained on a dataset created from 13,975 images (collected and compiled from 13,870 patients). Projection and expansion designs have been implemented. In this paper, CNN models VGG and ResNet achieve an accuracy of 83% and 90.6% respectively, while COVID-Net achieved 93.3% accuracy. Zhang et al. [16] use deep learning by using the X-Ray image dataset from a GitHub repository. The developed models show a sensitivity of 96% for COVID-19-positive cases and a sensitivity of 70.65% for COVID-19-negative cases. Tabik et al. [17] use a dataset known as the COVIDGR dataset which implemented the COVID-SDNet model. Here, the dataset has been created by collaborating with a hospital. The model achieves results of 97.72% (severe), 86.90% (moderate), and 61.80% (mild) accuracy in different severity levels. Abbas et al. [18] performed COVID-19 diagnosis using the DeTrac CNN model. The deep learning technique transfer learning has also been used. The model can deal with irregularities and achieved a very high accuracy of around 93.1%. Alghamdi et al. [19] performed a survey about using Deep Learning and CNNs for disease detection from Chest X-Rays. The study highlights the necessity of diverse datasets which should be publicly available. The common CNN models that were popular for experimentation among researchers are ResNet, DenseNet, GoogleNet/Inception,

COVID-19 Detection from Chest X-Ray Images

291

and VGGNet. The work in [21–25] focuses different techniques those deployed mainly image analysis techniques.

3 System Architecture and Design Table1 highlights the proposed framework of the study, showing layers and types of outputs shape and the total count of parameters used. The model contains a total of 20 layers where there are 6 are Convolutional, 4 of them are Max Pooling, and 6 Dropout. Then Flatten Layers and Dense Layer have also been used for the final predictions of the model. Table 1. Model architecture of proposed model Layer

Output shape

Convolution

(None, 329, 329, 32)

Parameters 896

Convolution

(None, 327, 327, 64)

18,496

Max pooling

(None, 163, 163, 64)

0

Dropout

(None, 163, 163, 64)

0

Convolution

(None, 161, 161, 64)

36,928

Max pooling

(None, 80, 80, 64)

0

Dropout

(None, 80, 80, 64)

0

Convolution

(None, 78, 78, 128)

73,856

Max pooling

(None, 39, 39, 128)

0

Dropout

(None, 39, 39, 128)

0

Convolution

(None, 37, 37, 128)

147,584

Max pooling

(None, 18, 18, 128)

0

Dropout

(None, 18, 18, 128)

0

Convolution

(None, 16, 16, 128)

147,584

Max pooling

(None, 8, 8, 128)

0

Dropout

(None, 8, 8, 128)

0

Flatten

(None, 8192)

Dense

(None, 64)

Dropout

(None, 64)

Dense

(None, 2)

0 524,352 0 130

Total parameters

949,826

Trainable parameters

949,826

Non-trainable parameters

0

Table 1 shows the layers used for the model. In addition to the self-designed model, the dataset has experimented on multiple built-in Keras CNN models such as VGG16,

292

N. B. Moin et al.

VGG19, Xception, ResNet101V2, ResNet152V2, InceptionV3, InceptionResNetV2, MobileNetV2, DenseNet201, NASNetLarge, and EfficientNetB3. The models used the Batch Normalization layers (renorm = True) and Global Average Pooling2D layers. Then, layers such as Dense Layers and Dropouts were used. The activations Relu and Softmax were used for the dense layers. 3.1 Dataset Description The dataset for disease detection dataset was obtained and collected from the data science platform Kaggle [20]. The dataset contains three directories where the images from the training and validation dataset were used for the study to train the models. The dataset contains two class labels, one for diseased lung X-rays and another for healthy X-ray images. The training and validation files have an equal distribution of images. In this dataset, a total of 348 images have been used as the inputs for the proposed method and the experimental CNN models. Figure 1 displays some of the sample images for both healthy and diseased X-Ray images from the training directory of the dataset.

Fig. 1. Sample input images data set (train)

COVID-19 Detection from Chest X-Ray Images

293

Table 2 shows detailed information about the distribution of images throughout the dataset for both Covid and Normal class labels in the case of both training and validation image datasets. Table 2. Image distribution of dataset Train/validation set

Class labels

Total images

Train dataset

Covid

144

Normal

144

Validation dataset

Covid Normal

Total

30 30 348

In total, 348 image samples were collected and used for analysis and experimentation. All the images are in png format in this open-sourced dataset. Total Covid images are 174 and healthy Chest X-Ray images are 174. A completely balanced distribution of images has been observed for the collected dataset as there is an equal number of images for both Covid and Normal samples. 3.2 Data Preprocessing In the proposed method for this study, a total of 348 images of both class labels have been used. For preprocessing the dataset and in order to make it suitable for training, the images have been resized where the input images have a size of 331 × 331 pixels. The images were rescaled. Data augmentation using the ImageDataGenerator function of Keras has been performed on the training and validation images. Horizontal and vertical flips have been applied after rescaling and the selected color mode was RBG. The training and validation split of the dataset was already created and separated into respective directories of the dataset.

4 Implementation and Experimental Result 4.1 Experimental Setup The experimentations were performed using the online platform Google Colaboratory. Google Colab contains the required latest version of Python along with additional dependencies and libraries such as Keras and Tensorflow. Keras has been used to create the CNN models and perform all the experimentations. The total number of epochs is 25, with a batch size equal to 8. Adam was the optimizer and cross-entropy (categorical) was the loss function. The chosen metric for evaluating the model performance for each epoch was the accuracy measure. The callback functions (early-stopping) with patience and restoring best weights have also been implemented into the model. The models were evaluated on all the images in training, validation, and testing.

294

N. B. Moin et al.

4.2 Performance Evaluation In order to accurately and comprehensively evaluate the efficiency of the system and all the experimented CNN models, the accuracy has been calculated using the following equations All the calculated values were obtained automatically from the Keras model training period. In addition to the accuracy, the calculation of the loss, precision, recall, and f1-score have also been performed for model evaluation. Accuracy =

No. of corrected correspondence × 100% No. of correspondence

(1)

True Positives True Positives + False Positives

(2)

Precision = Recall =

True Positives True Positives + False Negatives

F1 − score =

2 × (Precision × Recall) Precision + Recall

(3) (4)

The metrics and evaluation scores with the equations show how well the model is performing along with how good the model is for real-world experimentations and implementation. Table 3 shows that the accuracy of the proposed framework surpasses many previous works and is around 96.77%. The proposed framework gives a very low loss where the average loss is approximately 0.10429. Additionally, a high value of the other metrics can be observed which exceeds more than 93%. The obtained f1-score for the proposed system is 95%. In the case of precision, Covid-19 positive cases received a percentage of 94% and negative cases had 97%. While for recall, Covid-19 positive case is 97% and the negative case is 93%. Figure 2 shows the accuracies while Fig. 3 shows the losses over 25 epochs in both train and validation sets. The proposed framework shows an exponential increase in accuracy and a slow and smooth decrease in loss over the epochs. The graphs plotting the accuracy and loss for the epochs show an exponential curve. The model generalizes and learns from the data without errors. Along with the proposed model results, the results for the experimental built-in Keras models are briefly discussed in the following tables highlighting the accuracy, loss, etc. In Table 4, almost all the CNN models give a very high accuracy except EfficentNetB3. Out of all the models, VGG16, ResNet101V2, InceptionV3, and MobileNetV2 give the highest accuracy of 99%. Almost all the models here give higher accuracy than the proposed framework. VGG19, ResNet152V2, InceptionResNetV2, and NASNetLarge have the second-highest values of 97%. The most underperforming CNN model is the EfficientNetB3 having less than 50% accuracy. There was a decrease in accuracy percentages for the models when using validation data. The validation accuracies are similar to the training accuracies. Table 5 shows the loss calculated for all the models during training, validation, and evaluation. The lower the value of the model loss, the more efficiently the model is performing and the less prone to show errors. Low loss determines how the model

COVID-19 Detection from Chest X-Ray Images

295

Table 3. Performance analysis for proposed system Train accuracy

97.22%

Train loss

0.09587

Validation accuracy

96.42%

Validation loss

0.09375

Validation evaluation accuracy

96.67%

Validation evaluation loss

0.12327

Precision (COVID)

0.94

Precision (normal)

0.97

Recall (COVID)

0.97

Recall (normal)

0.93

F1-score (COVID)

0.95

F1-score (normal)

0.95

Fig. 2. Training and validation loss for proposed framework

Fig. 3. Training and validation accuracy for proposed framework

generalizes per epoch and improves performance and learning. The table shows that

296

N. B. Moin et al. Table 4. Performance analysis for built-in Keras CNN models

Models

Train accuracy (%)

Validation accuracy (%)

Validation evaluation accuracy (%)

VGG16

99.65

100.00

98.33

VGG19

93.40

91.07

91.67

Xception

98.95

96.42

95.00

ResNet101V2

99.30

98.21

96.67

ResNet152V2

97.22

94.64

95.00

InceptionV3

98.95

96.42

96.67

InceptionResNetV2

97.22

98.21

96.67

MobileNetV2

99.65

98.21

98.33

DenseNet201

98.26

96.42

98.33

NASNetLarge

96.87

96.42

96.67

EfficientNetB3

46.52

48.21

50.00

VGG16 and MobileNetV2 have the lowest loss values. Therefore, they are performing better compared to the other models. Xception, ResNet101V2, and InceptionV3 also have a comparatively lower loss, while ResNet152V2, InceptionResNetV2, DenseNet201, and NASNetLarge have comparatively higher loss values. EfficientNetB3 is the worstperforming model with a loss higher than 0.69. Table 5. Loss for built-in Keras CNN models Models

Train loss

Validation loss

Validation evaluation loss

VGG16

0.15574

0.14944

0.17977

VGG19

0.27254

0.31001

0.30134

Xception

0.19411

0.23464

0.26204

ResNet101V2

0.18148

0.19645

0.22280

ResNet152V2

0.20594

0.25528

0.24951

InceptionV3

0.19447

0.23364

0.23008

InceptionResNetV2

0.23756

0.21918

0.24458

MobileNetV2

0.17348

0.19489

0.19274

DenseNet201

0.20210

0.23204

0.20017

NASNetLarge

0.23647

0.24278

0.23910

EfficientNetB3

0.69470

0.69357

0.69323

COVID-19 Detection from Chest X-Ray Images

297

Table 6 shows the values obtained for all the mentioned metrics in the case of all the models. The values were obtained from the validation data. It can be seen that the scores for all the models and the 20-layered framework are comparatively higher than the required benchmark scores for this evaluation. Xception, ResNet152V2, and InceptionResNetV2 give comparatively higher scores where the values range from 94 to 100%. EfficientNetB3 gives the worst results with an f1-score of 67%. Table 6. Precision, recall, and F1-score for built-in models Models

Precision (C)

Precision (N)

Recall (C)

Recall (N)

F1-score (C)

F1-score (N)

VGG16

0.94

1.00

1.00

0.93

0.97

0.97

VGG19

1.00

0.94

0.93

1.00

0.97

0.97

Xception

0.97

1.00

1.00

0.97

0.98

0.98

ResNet101V2

0.97

1.00

1.00

0.97

0.98

0.98

ResNet152V2

0.88

1.00

1.00

0.87

0.94

0.93

InceptionV3

0.86

1.00

1.00

0.83

0.92

0.91

InceptionResNetV2

0.97

1.00

1.00

0.97

0.98

0.98

MobileNetV2

0.97

1.00

1.00

0.97

0.98

0.98

DenseNet201

0.94

1.00

1.00

0.93

0.97

0.97

NASNetLarge

0.97

1.00

1.00

0.97

0.98

0.98

EfficientNetB3

0.00

0.50

0.00

1.00

0.00

0.67

Tables 3 through 6 highlight the experimental results for all the CNN models and self-designed novel proposed framework along with a comparison of which models performed the best and which models performed the worst for the identification of the disease from the dataset images. 4.3 Comparison with Other Existing Frameworks Table 7 given below highlights the accuracies in other existing frameworks in contrast to the proposed framework and the built-in models with the previous works for both self-designed models and well-established CNN models. From the comparison, it can be said that most of the experimental models are showing high accuracies compared to the previous related works. VGG16, MobileNetV2, and ResNet101 show better performances compared to the previous works. The proposed framework’s accuracy does not exceed the previously designed frameworks but there is a very slight difference between the accuracies of the models.

298

N. B. Moin et al. Table 7. Comparison with previous works

Paper No.

Previous works results (%)

Previous works models

Experimented models performance (%)

[5]

97.00

Self-designed

96.77

[7]

90.13

VGG16

99.32

[7]

85.98

ResNet101

98.06

[8]

85.26

VGG16

99.32

[8]

87.37

ResNet101

98.06

[8]

89.47

VGG19

92.04

[9]

99.10

Self-designed

96.77

[10]

95.83

VGG16

99.32

[10]

95.83

MobileNetV2

98.73

[10]

92.50

InceptionV3

97.34

[15]

83.00

VGG19

92.04

[15]

93.30

Self-designed

96.77

5 Conclusion In this study, 11 popular Convolutional Neural Networks such as VGG16, VGG19, Xception, ResNet101V2, ResNet152V2, InceptionV3, InceptionResNetV2, MobileNetV2, DenseNet201, NASNetLarge, EfficientNetB3 were used. In addition to the eleven CNN models, a self-designed framework has been. All the models except EfficicentNetB3 give high accuracy and low loss. Most of the models outperform the previous works. The framework designed for this work gives an accuracy of 96.77% with a loss of 0.10429. In conclusion, the proposed framework and the built-in Keras models, especially VGG16, MobileNetV2, and ResNet101V2 give excellent performance in the damaged and diseased lungs from Chest X-Ray images for Covid. All those models can be implemented and tested on different datasets to compare and improve performances. From the eleven experimental and built-in Keras models used for this study, VGG16, MobileNetV2, and ResNet101V2 outperform the rest of the models. The proposed framework carries the lowest loss value out of all the models. Limitations to this study are fewer data usage for training. More data along with additional class labels for Pneumonia and Tuberculosis will further improve the prediction range and working range. Additionally, various techniques for transforming images can be applied to improve performance. These proposed techniques may also improve the precision, recall, and f1 scores. Additionally, kit tests can also be used together with this method for the accuracy of results. Future work for this study would include creating a web interface or a mobile application where the models will be used to conduct predictions in real time and as an end to end models.

COVID-19 Detection from Chest X-Ray Images

299

References 1. WHO: WHO coronavirus disease (COVID-19). https://www.who.int/health-topics/corona virus. Last accessed 21 May 2022 2. Rahaman, M.M., et al.: Identification of COVID-19 samples from chest X-ray images using deep learning: a comparison of transfer learning approaches. J. X-Ray. Sci. Technol. 28, 821–839 (2020). https://doi.org/10.3233/XST-200715 3. Moin, N.B., Islam, N., Sultana, S., Chhoa, L.A., Ruhul Kabir Howlader, S.M., Ripon, S.H.: Disease detection of Bangladeshi crops using image processing and deep learning—a comparative analysis. In: 2022 2nd International Conference on Intelligent Technologies. CONIT 2022, pp. 1–8 (2022). https://doi.org/10.1109/CONIT55038.2022.9847715 4. Sarvamangala, D.R., Kulkarni, R.V.: Convolutional neural networks in medical image understanding: a survey. Evol. Intel. 15(1), 1–22 (2021). https://doi.org/10.1007/s12065-020-005 40-3 5. Panwar, H., Gupta, P.K., Siddiqui, M.K., Morales-Menendez, R., Singh, V.: Application of deep learning for fast detection of COVID-19 in X-rays using nCOVnet. Chaos Solitons Fractals 138, 109944 (2020). https://doi.org/10.1016/j.chaos.2020.109944 6. Hasan Jahid, M., Alom Shahin, M., Ali Shikhar, M.: Deep learning based detection and segmentation of COVID-19 pneumonia on chest X-ray image. In: 2021 International Conference on Information and Communication Technology for Sustainable Development. ICICT4SD 2021—Proceedings, pp. 210–214 (2021). https://doi.org/10.1109/ICICT4SD50815.2021.939 6878 7. Basu, S., Mitra, S., Saha, N.: Deep learning for screening COVID-19 using chest X-ray images. In: 2020 IEEE Symposium Series on Computational Intelligence. SSCI 2020, pp. 2521–2527 (2020). https://doi.org/10.1109/SSCI47803.2020.9308571 8. Ismael, A.M., Sengür, ¸ A.: Deep learning approaches for COVID-19 detection based on chest X-ray images. Expert Syst. Appl. 164 (2021). https://doi.org/10.1016/j.eswa.2020.114054 9. Hussain, E., Hasan, M., Rahman, M.A., Lee, I., Tamanna, T., Parvez, M.Z.: CoroDet: a deep learning based classification for COVID-19 detection using chest X-ray images. Chaos Solitons Fractals 142, 110495 (2021). https://doi.org/10.1016/j.chaos.2020.110495 10. Nayak, S.R., Nayak, D.R., Sinha, U., Arora, V., Pachori, R.B.: Application of deep learning techniques for detection of COVID-19 cases using chest X-ray images: a comprehensive study. Biomed. Signal Process. Control. 64, 102365 (2021). https://doi.org/10.1016/j.bspc. 2020.102365 11. Heidari, M., Mirniaharikandehei, S., Khuzani, A.Z., Danala, G., Qiu, Y., Zheng, B.: Improving the performance of CNN to predict the likelihood of COVID-19 using chest X-ray images with preprocessing algorithms. Int. J. Med. Inform. 144, 104284 (2020). https://doi.org/10. 1016/j.ijmedinf.2020.104284 12. Minaee, S., Kafieh, R., Sonka, M., Yazdani, S., Jamalipour Soufi, G.: Deep-COVID: predicting COVID-19 from chest X-ray images using deep transfer learning. Med. Image Anal. 65 (2020). https://doi.org/10.1016/j.media.2020.101794 13. Mangal, A., Kalia, S., Rajgopal, H., Rangarajan, K., Namboodiri, V., Banerjee, S., Arora, C.: CovidAID: COVID-19 Detection Using Chest X-Ray, pp. 1–10 (2020) 14. Alazab, M., Awajan, A., Mesleh, A., Abraham, A., Jatana, V., Alhyari, S.: COVID-19 prediction and detection using deep learning. Int. J. Comput. Inf. Syst. Ind. Manag. Appl. 12, 168–181 (2020) 15. Wang, L., Lin, Z.Q., Wong, A.: COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-ray images. Sci. Rep. 10, 1–12 (2020). https://doi.org/10.1038/s41598-020-76550-z

300

N. B. Moin et al.

16. Zhang, J., et al.: Viral pneumonia screening on chest X-rays using confidence-aware anomaly detection. IEEE Trans. Med. Imaging 40, 879–890 (2021). https://doi.org/10.1109/TMI.2020. 3040950 17. Tabik, S., et al.: COVIDGR dataset and COVID-SDNet methodology for predicting COVID19 based on chest X-ray images. IEEE J. Biomed. Heal. Inform. 24, 3595–3605 (2020). https:// doi.org/10.1109/JBHI.2020.3037127 18. Abbas, A., Abdelsamea, M.M., Gaber, M.M.: Classification of COVID-19 in chest X-ray images using DeTraC deep convolutional neural network. Appl. Intell. 51(2), 854–864 (2020). https://doi.org/10.1007/s10489-020-01829-7 19. Alghamdi, H.S., Amoudi, G., Elhag, S., Saeedi, K., Nasser, J.: Deep learning approaches for detecting COVID-19 from chest X-ray images: a survey. IEEE Access. 9, 20235–20254 (2021). https://doi.org/10.1109/ACCESS.2021.3054484 20. Fenta, F.: Chest X-ray for covid-19 detection. https://www.kaggle.com/datasets/fusicfenta/ chest-xray-for-covid19-detection. Last accessed 15 May 2022 21. Saha, R., Debi, T., Arefin, M.S.: Developing a framework for vehicle detection, tracking and classification in traffic video surveillance. In: Vasant, P., Zelinka, I., Weber, GW. (eds.) Intelligent Computing and Optimization. ICO 2020. Advances in Intelligent Systems and Computing, vol. 1324. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-681548_31 22. Fatema, K., Ahmed, M.R., Arefin, M.S.: Developing a system for automatic detection of books. In: Chen, J.I.Z., Tavares, J.M.R.S., Iliyasu, A.M., Du, K.L. (eds.) Second International Conference on Image Processing and Capsule Networks. ICIPCN 2021. Lecture Notes in Networks and Systems, vol. 300. Springer, Cham (2022). https://doi.org/10.1007/978-3-03084760-9_27 23. Rahman, M., Laskar, M., Asif, S., Imam, O.T., Reza, A.W., Arefin, M.S.: Flower recognition using VGG16. In: Chen, J.I.Z., Tavares, J.M.R.S., Shi, F. (eds.) Third International Conference on Image Processing and Capsule Networks. ICIPCN 2022. Lecture Notes in Networks and Systems, vol. 514. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-12413-6_59 24. Yeasmin, S., Afrin, N., Saif, K., Imam, O.T., Reza, A.W., Arefin, M.S.: Image classification for identifying social gathering types. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). https://doi.org/10.1007/ 978-3-031-19958-5_10 25. Ahmed, F., et al.: Developing a classification CNN model to classify different types of fish. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). https://doi.org/10.1007/978-3-031-19958-5_50

A Note on Solving the Transportation Model by the Hungarian Method of Assignment: Unification of the Transportation and Assignment Models Santosh Kumar1 , Trust Tawanda2 , Elias Munapo3(B) , and Philimon Nyamugure2 1 Department of Mathematical and Geospatial Sciences, School of Sciences, RMIT University,

Melbourne, Australia [email protected] 2 Department of Statistics and Operations Research, National University of Science and Technology, Bulawayo, Zimbabwe {trust.tawanda,philimon.nyamugure}@nust.ac.zw 3 School of Economics and Decision Sciences, North West University, Mafikeng Campus, Mafikeng, South Africa [email protected]

Abstract. This short note extends the Hungarian method for the assignment for solving a transportation model. In the proposed approach, the optimality of the solution is established when the solution is feasible, hence a degenerate transportation solution poses no difficulties as is the case of the conventional transportation approach. Keywords: Hungarian method · Assignment · Transportation · Degeneracy

1 Introduction The assignment and transportation models are well-known in the OR literature, and they have many industrial applications in production planning, telecommunication, scheduling, military operations etc., see Hillier and Lieberman [1], Munapo and Kumar [2], Taha [3], Tawanda [4] and Winston [5]. Both models are essentially degenerate linear programs, hence solution by the simplex approach was considered inefficient for solving these two models. In search for solving these two special models, special methods were developed to solve the assignment and transportation models, which have been well documented in OR books, for example, Taha [3] and Winston [5]. Models for both problems are totally uni-modular, hence these two LP models result in an integer solution. Solution of the assignment model was proposed by Kuhn [6] and he called his approach, ‘The Hungarian method for assignment’. The method by Kuhn uses some properties associated with the cost matrix for the assignment problem and developed an algorithm that gets to a solution which requires no proof of optimality; hence the approach is free of difficulties associated with a degenerate LP solution. However, for © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 301–313, 2024. https://doi.org/10.1007/978-3-031-50158-6_30

302

S. Kumar et al.

the transportation model, a special method was developed that used the properties of the dual of the transportation LP and developed an approach that requires an initial feasible solution, a testing procedure for optimality and if the solution does not satisfy optimality conditions, a process to move to a better solution. This process is repeated, until the optimal solution is reached. Efficiency of this approach is dependent on the initial starting solution and the number of basic variables, which is required to be (m + n −1), where m denotes the number of rows and n denotes the number of columns in the transportation model. Therefore, methods were developed to find an initial solution, which was subjected to an optimality test and if the solution failed the test for optimality, iteratively the solution was improved until it satisfied the optimality test. Recently the ‘method of subtractions’ was developed and is presented in Munapo and Kumar [2]. For a degenerate solution, computational load for the test of optimality increases and makes it more difficult to establish the optimality of a solution. Degeneracy in LP or for this special class of model can cause difficulties as the value of the objective function may not improve in successive iterations. In many instances, one may have reached the optimal solution, but the test of optimality is unable to recognise its optimality due to degeneracy. The assignment and transportation are both degenerate with respect to LP model and order of degeneracy for the transportation problem as a LP model is 1 or more and for the assignment model, degeneracy order is n. In this short note, without any loss of generality, a balanced transportation problem is considered as an assignment model, and the Hungarian method of assignment is modified to solve the transportation model. Although the unification of transportation and assignment was established earlier by Munapo et al. [7], but in this note the Hungarian method is slightly modified to solve the transportation model also. The paper has been organized in 5 sections. Mathematical models and necessary background have been discussed in Sect. 2. In this section, the modified Hungarian method for the transportation problem is also presented. Some interesting properties are presented in Sect. 3. Three numerical illustrations have been presented in Sect. 4, and finally some concluding remarks have been discussed in Sect. 5.

2 Mathematical Models and Necessary Background First, we recall the mathematical models of both the assignment and the transportation models in Sect. 173.1 and relook at the transportation model as an assignment model in Sect. 173.2. 2.1 Mathematical Formulations of Transportation and Assignment Models The mathematical model of a (m × n) transportation model is given by: Min Z = n  j=1

n m  

Cij xij

i=1 j=1 m 

xij = ai ,

i=1

xij = bj , i = 1, 2, . . . , m; j = 1, 2, . . . , n.

(1)

A Note on Solving the Transportation Model

303

Further, without any loss of generality, it is assumed that the problem is a balanced transportation problem, i.e. m 

ai =

i=1

n 

bj = M

j=1

xij ≥ 0 and an integer for all i and j

(2)

Note Cij denotes the per unit cost of transportation of a unit from warehouse i to the destination j, ai and bj denote supply at the warehouse i and demand at the destination j where i = 1, 2, . . . , m and j = 1, 2, …, n. Similarly, the mathematical model of an (m by n, where m = n) assignment problem is given by (3) Min Z = n  j=1

n m  

Cij xij

i=1 j=1 m 

xij = 1,

xij = 1, i = 1, 2, . . . , m; j = 1, 2, . . . , n, m = n

i=1

xij = 0 or 1 for all i and j.

(3)

For an assignment problem, all ai = bj = 1 and m = n. 2.2 Consideration of the Transportation Model as an Assignment Model If each row i in the transportation model is repeated for ai number of times and similarly each column j is repeated for bj number of times, each element Cij in the transportation cost matrix can be seen as a submatrix of dimension ai and bj , as shown below: ⎡

⎤ Cij,11 . . . Cij,1bj ⎣ ... ... ... ⎦ Cij,ai1 . . . Cij,aibj When each C ij for all i and j is expanded in the form of a matrix as given by the matrix above, the new problem will be a M × M size assignment problem n with demand  a = and supply 1 at each supply and demand point, Here M = m i i=1 j=1 bj . Since the new M × M size problem will be an assignment model, it is natural that the Hungarian method of assignment will be applicable and unlike the conventional transportation method, the problem can be solved as an assignment problem by using the Hungarian method of assignment. First, we make some observations in Sect. 173.3 and develop the Hungarian method for the transportation problem in Sect. 173.4.

304

S. Kumar et al.

2.3 Some Observations Observation 1 The Hungarian approach develops equivalent cost matrices and attempts to find the assignment at zero total cost which is a natural minimum, hence optimality proof is not required. Observation 2 The Hungarian method of assignment, deals with the modified cost elements and after detecting an independent zero element, makes an allocation and draws a horizontal or a vertical line, to indicate no further assignments are possible in that row or column. However, in the case of transportation problem, if the requirement of such independent zero elements is represented by a number p, then the number p must satisfy the condition (4). Max (m, n) ≤ p ≤ (m + n − 1)

(4)

Observation 3 In a non-degenerate balanced transportation problem, the number of allocations will be given by (m + n − 1). It means the number of horizontal lines must not exceed (m − 1) and similarly the number of vertical lines also must not exceed (n − 1) for further modification of the relative cost elements Cij . This observation is obvious, as when all elements are on a line, no element will be left as a positive element, which acts like a seed for further modification of the cost elements Cij . Observation 4 In a non-degenerate balanced transportation problem, maximum number of independent zero elements that can be created by the Hungarian method of assignment can be at most (m + n − 1), which is exactly the requirement for determination of the optimal solution of the transportation model. Hence the Hungarian method of assignment can be extended to sole the balances transportation model. 2.4 Modified Hungarian Method of Assignment for Solving Transportation Problem The steps of the Hungarian Method of assignment for the balanced transportation model will be as follows: Step 1. Consider the balanced transportation model, as all transportation models can be easily converted to a balanced model by the introduction of an appropriate dummy supply or a demand point. Step 2. Subtract the minimum element in each row from all other elements in that row. This process will generate at least one zero element in each row. Repeat the same process for each column, thus once again, each column will have at least one zero element in each column in the given balanced (m × n) transportation model.

A Note on Solving the Transportation Model

305

Step 3. Inspect row i, where i = 1, 2, . . . , m, if the row i has a unique zero element, i.e. the modified Cij = 0 in column j, where j = 1, 2, . . . , n; find the min ai , bj . (1) If ai = bj , draw a vertical line through the column j, as was done in the case of an assignment model. The modified value of ai will be zero and similarly the modified value of bj will also be zero. In other words, the (m × n) model will be reduced to ((m − 1) × (n − 1)) model. (2) If ai ≤ bj , draw a horizontal line through the row i, if the number of horizontal lines is (m − 1). Change the value ai to zero, and the bj value is changed to  less than bj − ai . (3) if ai ≥ bj , draw a vertical line in column j, if the number of vertical lines is less that (n − 1). Change the value of bj to zero, and the value of ai will be changed to ai − bj . Repeat the same process for each row and column, and when all zero elements have been covered, go to Step 4. If all zero elements have not been covered by a horizontal or a vertical line, find for each zero the possible maximum allocation and draw the line as per the rule described (1) to (3). Repeat this process until all zero elements have been covered. Go to Step 4. Step 4. If the solution is feasible, go to Step 6, and if the solution is not feasible, and all zero elements in the modified matrix have been covered, do the following: If the number of lines are less than (m + n − 2), find the minimum positive modified element, which have not been crossed out by a vertical or the horizontal line. Using the value of this element, update the cost matrix as follows: (i). Subtract this minimum value form all elements which have not been crossed out. (ii). Do not alter values which are on a single line. (iii). Add the minimum value to all elements which are on the intersection of two lines. If the number of lines are equal to (m + n − 2), only one element will be left as uncrossed out. If this number is large compared to other elements in that row or column, we get the feasible solution, and go to Step 7. Step 5. Go to Step 2, i.e. make sure once again that each row and column has a zero. Step 6. Each allocation results in a horizontal or a vertical line, and if total number of lines is equal to (m + n − 2), then the current allocation must lead to a feasible solution, and that will be an optimal solution, if this is minimum in that column or row, as may be the case. The process will terminate. Step 7. If the last element is not minimum, obtain a feasible solution, and subject it to the test of optimality, and obtain the optimal solution by the conventional approach. This process will be able to handle degenerate transportations models, without any difficulty.

306

S. Kumar et al.

3 Interesting Properties of a Transportation Model 3.1 Property 1 Unlike a linear program, a special characteristic of a balanced transportation model is that any feasible solution can be tested for optimality without any past information. Proof The LP approach is sequential, for example, the extreme point search at any point of the LP convex space is dependent on the previous extreme point, in other words, if a LP is given and its corresponding optimal solution is given, we must resolve to prove the optimality of the given solution. In the case of a transportation model, the optimality proof is based on values of the dual variables of the transportation model, which are directly linked to the given basic variables. Therefore, given a basic feasible solution, it can be tested for optimality at any time. This the property we use in Step 7. 3.2 Property 2 The Hungarian approach discussed in this paper deals with infeasible solutions, which is partly situated at zero modified cost elements and partly it will be at non-zero positive modified cost elements. Hence one can easily move from the Hungarian approach to conventional transportation approach to test the optimality of the solution at any stage of iterations. However, the Hungarian approach is sequential similar to the LP i.e. the process will have to start from the very beginning, if previous information is missing.

4 Numerical Illustrations We present three illustrations, dealing with alternative, unique and degenerate solutions. 4.1 Illustration 1 Consider an example of a transportation problem taken from Eppen et al. [8, p. 283]. The cost per unit, demand and supply is given in Table 1. Table 1. The transportation problem 1

2

3

4

Supply

1

12

13

4

6

500

2

6

4

10

11

700

3 Demand

10

9

12

4

800

400

900

200

500

2000

Note the minimum element in each row is 4. For creating a zero in each row, we subtract 4 from each element, we obtain Table 2 as given below.

A Note on Solving the Transportation Model

307

Table 2. Modified values of the unit transportation cost 1

2

3

4

Supply

1

8

9

0

2

500

2

2

0

6

7

700

3

6

5

8

0

800

400

900

200

500

2000

Demand

Since column 1 does not have a zero element, we again subtract 2 from the elements in column 1, the costs per unit are further modified as give in Table 3. Allocated values are shown in parenthesis and horizontal lines through a row are indicated by → and a vertical line through a column is indicated ↑ in the supply and demand column respectively. A number with an arrow indicates the order of the line. For example, in Table 3, column 3 has a vertical arrow with a number 1, it means first allocation was made in this column in the cell (1, 3). Table 3. Modified values of the unit transportation cost 1

2

3

4

Supply

1

6

9

0 (200)

2

500, 300

2

0 (400)

0 (300)

6

7

700, 300, 0 → 4

3

4

5

8

0 (500)

800, 300

Demand

400, 0 ↑ 3

900, 600

200, 0 ↑ 1

500, 0 ↑ 2

2000, 600

Note that columns 1, 3, and 4 have been crossed out by vertical lines and row 2 by a horizontal line and the solution is not yet feasible, as 600 more units must be allocated. We need to create more zero elements. Among the two not crossed out elements in column 2, the minimum is 5, situated in the cell (3, 2). The updated values are given in Table 4. Table 4. Modified values of the unit transportation cost from Table 3 1

2

3

4

Supply

1

6

4

0

2

500

2

5

0

11

12

700

3

4

0

8

0

800

400

900

200

500

2000

Demand

308

S. Kumar et al.

Note that column 1 has no zero elements, hence subtracting the minimum, a zero element is created as shown in Table 5. Table 5. Modified values of the unit transportation cost 1

2

3

4

Supply

1

2

4

0 (200)

2

500, 300

2

1

0 (700)

11

12

700, 0 → 2

3

0 (400)

0 (200)

8

0 (200)

800, 400, 200 0→5

Demand

400, 0 ↑ 3

900, 2000 ↑ 4

200, 0 ↑ 1

500, 300

2000, 300

The solution is still infeasible, hence more zeros must be created. Columns 1, 2 and 3 have been covered by vertical lines, rows 2 and 3 by the horizontal lines. Hence the only element not crossed out is in the cell (1, 4). Its value is 2, which is minimum element in row 1. The new updated values are shown in Table 6. Table 6. Modified values of the unit transportation cost 1

2

3

4

Supply

1

2

4

0

0

500

2

3

2

13

12

700

3 Demand

2

2

10

0

800

400

900

200

500

2000

Now columns 1 and 2 have no zero elements, subtracting the minimum, and making allocations, we obtain Table 7. Table 7. Modified values of the unit transportation cost from Table 6 1

2

3

4

Supply

1

0

2

0 (200)

0 (300)

500, 300, 0

2

1

0 (700)

13

12

700, 0, 0 → 1

3

0 (400)

0 (200)

10

0 (200)

800, 600, 200, 0

Demand

400, 0 ↑ 4

900, 200, 0 ↑ 2

200, 0 ↑ 3

500, 300, 0

2000, 0

In Table 7, we have two zero elements in row 1 and 3 and two zeros in column 1 and 4, hence more than one optimal solution may exist. One possible solution is displayed

A Note on Solving the Transportation Model

309

in Table 7, which is optimal and total cost is $12,000. The second alternate solution is shown in Table 8. Table 8. The alternate optimal solution 1

2

3

4

Supply

1

12 (300)

13

4 (200)

6

500

2

6

4 (700)

10

11

700

3

10 (100)

9 (200)

12

4 (500)

800

Demand

400

900

200

500

2000

4.2 Example 2 Consider one more transportation problem taken from Munapo and Kumar [2, p. 132] as given below in Table 9. Table 9. The transportation model 1

2

3

4

4

Supply

8

9

2

2

1

1

1

1

1

30

3

2

8

19

7

3

14

40

4

2

6

6

4

11

8

50

Demand

4

6

1

5

20

5

6

20

2

15

5

12

8

7

60

30

40

50

20

20

40

200

After subtracting 4 from element of row 1, 1 from the elements in row 2, and 2 from the elements in row 3, 4 and 5. This results in Table 10. Note that each column in Table 10 has a zero element, hence no further modifications will arise in the modified cost elements. Following the allocation procedure, we get allocations as shown in Table 11. The solution in Table 11 is not feasible, hence we must update the cost elements in Table 11 following the Step 4, as shown in Table 12. The minimum non-zero element is 1, which has not been crossed out. Note the horizontal lines are in row 1 and 2 and the vertical line is in column 1. The solution in Table 12 is still infeasible. The minimum uncrossed element is 1. The updated table is given in Table 13. Note the minimum uncrossed element is 1, which will be used to update the elements as shown in Table 14.

310

S. Kumar et al. Table 10. With at least one zero in each row 1

2

3

4

5

6

Supply

1

4

5

16

0

0

2

20

2

1

0

0

0

0

0

30

3

0

6

17

5

1

12

40

4

0

4

4

2

9

6

50

5

0

13

3

10

6

5

60

30

40

50

20

20

40

200

Demand

Table 11. Allocation in cells with modified cost 0 1

2

3

4

5

6

Supply

1

4

5

16

0 (20)

0

2

20, ---- 0 → 3

2

1

0 (30)

0

0

0

0

30, ---- 0, → 2

3

0 (30)

6

17

5

1

12

40, ---- 10

4

0

4

4

2

9

6

50 ----

5

0

13

3

10

6

5

60

Demand

30, 0 ↑ 1

40, ---- 10

50

20, ---- 0

20

40

200, ---- 120

Table 12. Updated elements as shown below 1

2

3

4

5

6

Supply

1

5

5

16

0 (20)

0

2

20, 0 → 3

2

2

0 (30)

0

0

0

0

30, ---- 0 → 2

3

0

5

16

4

0 (20)

11

40, ---- 20

4

0 (30)

3

3

1

8

5

50, ---- 20

5

0

12

2

9

5

4

60

Demand

30, 0 ↑ 1

40, ---- 10

50

20, ---- 0

20, ---- 0 ↑ 4

40

200, ---- 100

Note the minimum of those elements which have not been crossed out is 1, which is again minimum. The updated information is given in Table 15. The solution in Table 15 is feasible, hence optimal. Note that the optimal solution mentioned by Munapo and Kumar [2] is 780, which is the same as obtained by the modified Hungarian method for the transportation problem.

A Note on Solving the Transportation Model

311

Table 13. Updated elements from Table 12 as shown below 1

2

1

6

5

2

3

0 (30)

3

0

4

4

0 (30)

5

0

Demand

30, 0 ↑ 2

3

4

5

6

Supply

16

0 (20)

1

2

20, 0

0

0

1

0

30, ---- 0 → 3

15

3

0 (20)

10

40, ---- 20

2

2

0

8

4

50, ---- 20

11

1

8

5

3

60

40, ---- 10

50

20, ---0↑1

20, ---0↑4

40

200, ---- 100

Table 14. Updated elements from Table 13 as shown below 1

2

3

4

5

6

1

6

4

2

4

0 (30)

3

0

3

14

4

0 (30)

1

5

0

10

Demand

30, 0 ↑ 2

40, ---- 10

Supply

15

0 (20)

1

1

20, 0

0

1

2

0

30, ---- 0 → 3

3

0 (20)

9

40, ---- 20

1

0

8

3

50, ---- 20

0 (50)

8

5

2

60, ---- 10

50 ---0↑5

20, ---0↑1

20, ---0↑4

40

200, ---- 50

Table 15. Updated elements from Table 6 as shown below 1

2

3

4

5

6

Supply

1

6

3

15

0 (10)

1

0 (10) 20, 10, ---- 0 → 9

2

5

0

1

2

3

0 (30) 30, 0 → 6

3

0 (20)

2

14

3

0 (20)

8

40, ---- 20, ---- 0 → 3

4

0

0 (40)

1

0 (10)

8

2

50, ---- 10, ---- 0 → 7

5

0 (10)

9

0 (50)

8

5

1

60, ---- 10, 0

Demand 30, 10 0 ↑ 4* 40, ---- 0 ↑ 5* 50, ---- 0 ↑ 1* 20, ---- 0 ↑ 8* 20, ---- 0 ↑ 2* 40, ---- 20 200, ------ 0

4.3 A Degenerate Transportation Model We now consider a degenerate transportation model taken from Sharma [9] as given in Table 16. After creating a zero element in row and column, we get from Tables 16 and 17, as given below with allocations.

312

S. Kumar et al. Table 16. Degenerate transportation model D1

D1

D1

Supply

S1

8

5

6

120

S1

15

10

12

80

S1

3

9

10

80

150

80

50

280

Demand

Table 17. With modifies per unit transportation cost D1

D1

D1

Supply

S1

3

0

0 (50)

120, 70

S1

5

0 (80)

1

80, 0

S1

0 (80)

6

6

80, 0 → 2

Demand

150, 70

80, 0 ↑ 1

50, 0, ↑ 3

280, 70

Since the solution is infeasible, more zero elements will have to be created. Vertical lines in column 2 and 3, and horizontal line is in row 3. The minimum of the element not crossed out is 3. The updated values are given in Table 18. Table 18. Updated values from Table 17 D1

D1

D1

Supply

S1

0 (70)

0

0 (50)

120, 70, 0, → 4

S1

2

0 (80)

1

80, 0

S1

0 (80)

9

9

80, 0 → 2

Demand

150, 70, 0

80, 0 ↑ 1

50, 0, ↑ 3

280, 70, 0

The solution in Table 18 is feasible, hence optimal. Once again, it is the same solution as given by Sharma [9]. Normally for this problem, the test for optimality will require 5 allocations, but by the Hungarian method for the transportation problem, degeneracy does not cause any difficulty for establishing the optimality of the solution.

5 Concluding Remarks In this short note, the Hungarian method of assignment has been extended to solve the transportation model. Optimality of the feasible solution is independent of the order of degeneracy. Several problems were solved, and they all resulted in the optimal solution. Testing optimality of a feasible is easy in the transportation model, and can be easily switched on for testing and optimality.

A Note on Solving the Transportation Model

313

In this paper, since transportation model has been considered as an assignment model, we can once again find the Kth best solution of the transportation model, where k ≥ 2. Please see, Murthy [10] and Kumar et al. [11]. In the case of a transportation model, if we have ‘p’ number of allocations, where the value of ‘p’ is governed by the Eq. (7), the 2nd best solution to the transportation model can be determined by approaches discussed by Murthy [10] and Kumar et al. [11] and Munapo and Kumar [2].

References 1. Hillier, F.S., Lieberman, G.J.: Introduction to Operations Research. ISBN 13:79812 59545962 (2015) 2. Munapo, E., Kumar, S.: Linear Integer Programming: Theory, Applications, Recent Developments, De Gruyter Series on the Applications of Mathematics in Engineering and Information Sciences. ISBN 978-3-11-070292-7 (2022) 3. Taha, H.A.: Operations Research: An Introduction, Pearson Educators, 10th edn. (2017) 4. Tawanda, T.: A node merging approach to the transhipment problem. Int. J. Syst. Assur. Eng. Manag. 8(Suppl 1), 370–378 (2017). https://doi.org/10.1007/s13198-015-0396-9 5. Winston, W.L.: Operations Research Applications and Algorithms, Duxbury Press, 4th edn. (2004) 6. Kuhn, H.W.: The Hungarian method of assignment problem. Naval Res. Log. Quart. 2, 83–97 (1955) 7. Munapo, E., Nyamugure, P., Lesaoana, M., Kumar, S.: A note on unified approach to solving transportation and assignment models in operations research. Int. J. Math. Model. Simul. Appl. 5, 140–149 (2012) 8. Eppen, G.D., Gould, F.J., Schmidt, C.P.: Introduction to Management Science, 4th edn. Prentice Hall, New Jersey (1993) 9. Sharma, J.K.: Operations Research: Theory and Applications. Trinity Press (2018) 10. Murthy, K.G.: An algorithm for ranking all the assignments in order of increasing costs. Oper. Res. 16, 682–687 (1968) 11. Kumar, S., Al-Hasani, A., Al-Rabeeah, M., Ebehard, A.: A random search method for finding ‘k ≥ 2’ number of ranked optimal solution to an assignment problem. J. Phys. Conf. Ser. 1490, 1–3 (2020). https://doi.org/10.1088/1742-6596/1490/1/012063

Automatic Crack Detection Approach for the Offshore Flare System Inspection Teepakorn Tosawadi1(B) , Pakcheera Choppradit1 , Satida Sookpong2 , Sasin Phimsiri1 , Vasin Suttichaya1 , Chaitat Utintu1 , and Ek Thamwiwatthana1 1 AI and Robotics Ventures Co.,Ltd., Bangkok, Thailand

[email protected] 2 Skyller Solutions Co., Ltd., Bangkok, Thailand

Abstract. Asset inspection using unmanned aerial vehicle (UAV) is gaining more attention in surveillance and exploratory engineering. UAV surveillance enables higher efficiency in terms of operating budget and time. This technique is usually used to gather asset information for detecting the anomalies, such as cracks and rust. However, cracking detection on the images from UAV’s camera also requires expertise to examine each image, which is time-consuming. To mitigate the issue, automatic crack detection on the flare stack images is proposed in this paper. This research used UAV with high-resolution cameras to collect the flare stack images from the oil refineries. The proposed method introduced crack detection using three object detection models, YOLOv5, YOLOv7, and DINO. Additionally, the Sobel operator and Canny edge detector are also applied to preprocessing images. With an inference time of 50 ms and performance of 65.4% mAP0.5 and 32.2% mAP0.5:0.95, the results show that YOLOv7-W6 on RGB images had the best performance and fastest inference time. Keywords: Flare stack inspection · Crack detection · Flare stack crack

1 Introduction In oil and gas industry assessment inspection, the flare stack of off-shore and on-shore oil refineries is one of the key components. The flare stack is used for discharging unwanted gas combustion and must be fully functional for ensuring safety in the petrochemical and refinery sectors. The failure can cause devastating damage to both human society and the environment. With this consequence, the flare stack is required to be inspected routinely. In recent years, UAVs with the high-resolution camera was deployed for inspection task widely. An inspection using UAVs benefits faster and safer inspection. Additionally, the refinery can operate normally during the inspection process. However, the above method requires inspector expertise, which is costly and consumes time. Each inspector also has unique skills and experience, that result in different analysis results on the same image. Additionally, expert fatigue and throughput can lead to error inspection. Therefore, the tools are required to set standardization and relieve inspector workload. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 314–323, 2024. https://doi.org/10.1007/978-3-031-50158-6_31

Automatic Crack Detection Approach for the Offshore Flare …

315

Many organizations are aware of crack detection. The earlier research failed to detect the crack in our dataset. The reason is the datasets used in many previous works were obtained from controlled environments. Moreover, most of them focus on concrete and road surfaces. For flare stack crack, there is no open dataset because of the difficulty of data collection. UAV with a camera is the most effective way to collect images of the flare stacks, but the dynamic scene will be an issue. Additionally, the large coverage area and high quality of the images from UAV will affect the model’s performance and efficiency since they increase the amount of model training and necessitate picture downscaling, which results in the loss of features from small objects. This paper investigates the most effective way to apply the object detection models to locate the cracks on the flare stack surface. Two image preprocessing techniques, including the Sobel operator and Canny edge detector, are applied to create an image emphasizing edges. We further investigate the effect of slicing the high-resolution image into several small images. Three object detection models, including YOLOv5, YOLOv7, and DINO, are trained to detecting cracks in images.

2 Related Works 2.1 Object Detection Methods DETR with Improved deNoising anchOr boxes (DINO) [1] was developed by The International Digital Economy Academy (IDEA). It was improved from two extensions of the transformer-based model, namely DAB-DETR [2] and DN-DETR [3] to increase performance and efficiency. The DINO model introduced two different backbones, namely Residual Network (Resnet) [4] and Swin transformer [5]. Finally, it achieved good performance with both small model size and tiny data size compared with other State-of-the-art models. You only look once (YOLO) [6] was developed to focus on real-time prediction by Joseph Redmon et al. Later that, the YOLO families such as YOLOv2 [7], YOLOv3 [8], YOLOv4 [9], YOLOv5 [10], YOLOv6 [11] and YOLOv7 [12] were published in order to enhance the original YOLO model. YOLOv5 was developed by Ultralytics. It suggested the spatial pyramid pooling (SPP) technique which improves the detection of small objects. It is also the most widely utilized in particle applications due to the model’s ability to export in a variety of formats, including TensorFlow-lite and Tensorrtengine. YOLOv7 is the most recent version of the YOLO family, which offers significant improvements through the use of focal loss and nine anchor boxes. As compared to other YOLO models, it achieved the fastest inference speed.

316

T. Tosawadi et al.

2.2 Crack Detection There are several approaches to detecting cracks in images. These techniques can be divided into two categories: conventional computer vision approach and deep learning approach [13–15]. Computer vision-based methods, such as thresholding, edge detectors, and matched filtering algorithm, have the disadvantage of being sensitive to light noise and requiring preprocessing and postprocessing. Therefore, many papers introduce the machine learning approach to detect cracks. Hafiz Suliman Munawar et al. applied Cycle Generative Adversarial Network (CycleGAN) [13] to detect the damages on roads, buildings, and bridges. Qin Zou et al. developed an end-to-end road crack segmentation method called DeepCrack [14] by a deep hierarchical convolutional neural network (CNN). Another road crack segmentation was introduced by Ju Huyan et al., namely CrackU-net [15]. It improved the other traditional method by the use of advanced CNN and “U”-shaped model architecture.

3 Methodology 3.1 Data Acquisition This research is conducted on our image dataset. The data were gathered from the offshore and on-shore oil refinery flare stack. An image is taken using a high-resolution camera attached to the UAV, operated by an experienced inspector. The data was taken under actual flare stack inspection operation where the light condition is dynamic and uncontrollable. Even though exposure and brightness are adjustable using camera settings, the standard policy remains unable to be exactly defined. Due to data scarcity and high operation costs, we can collect the data of 780 images in total. The data contains a set of images with 4:3 and 16:9 aspect ratios, and resolution varies from 1600 × 1200 to 6784 × 3816 pixels. The dataset composes of various operation sites, different flare stack categories, and a diversity of shot types. 3.2 Proposed Method In this work, crack detection using object detection is proposed to find a capability and suitable approach for flare stack inspection applications. The proposed method is illustrated in Fig. 1. We introduced a combination of three input types and three object detection models. Firstly, we fed raw RGB input images to each object detection model to produce a baseline experiment result. Then, we applied two edge detection techniques, including the Sobel operator and the Canny edge detector, to emphasize the edge of cracks. Typically, the RGB image is used as an input for deep learning object detection since it was developed to eliminate some complex handcrafted feature extraction. However, in some specific tasks like crack detection, adding preprocessing using traditional image processing techniques may enhance deep learning performance. Input images are processed before they were fed into deep-learning object detectors instead of RGB images. The Sobel operator is constructed using 3 × 3 kernels to find a gradient in both vertical and horizontal directions. Then, the gradient magnitude is combined to generate

Automatic Crack Detection Approach for the Offshore Flare …

317

Fig. 1. The overview of proposed method.

an image with edge feature enhancement. A Canny edge detector is developed on the Sobel operator with additional improvement. We also applied a Gaussian filter to blur the image in the Sobel operator. Applying the Gaussian filter is necessary since it resulted in edge detection with less noise. Since the object detection was pre-trained with threechannel data, preprocessed images are also converted from grayscale to three-channel RGB images. Regarding proposed object detection approaches, YOLOv5 [10] and YOLOv7 [12] are selected for CNN-based detectors. YOLOv5 is the most popular YOLO architecture, which was published by Ultralytic. Recently, YOLOv7 is the latest version proposed by Alexey Bochkovskiy, the original author of YOLO. There are many variant subsidiary models scaling based on architecture complexity, parameters, and input size. Since we want to preserve the quality and features of drone images, the network that is scaled for larger input image resolution is chosen. In the transformer-based architecture, we trained the detector using a state-of-the-art DINO algorithm. DINO was placed with two types of backbone which are ResNet101 and SwinTransformer Small.

4 Experimental Results and Discussion The experiment in this research was conducted to answer the following research questions: Q1: Which is an appropriate combination of preprocessing and object detection? Q2: Which approach gives the best result in terms of cost-accuracy compensation? Q3: Is the image-slicing approach suitable for crack detection on a multi-scaled dataset? 4.1 Data Preparation The image dataset is randomly split into the train, validation, and test subset. Our dataset consisted of 70% train images, 10% validation images, and 20% test images. The objects are annotated with one class category defined as crack. A random split is performed five times using different random seeds. Each model will be trained five times repeatedly, then averaging evaluation matrices to reduce performance fluctuation.

318

T. Tosawadi et al.

4.2 Experiment Setup In the proposed method, each algorithm has many variants of its based architecture and many fine-tuning options. The experiment must be conducted in a similar limited environment. Thus, the model capabilities are explored on similar standards and avoid a biased performance result comparison. All the experiments used pre-trained models. The model was trained on obtained input images with resolution 1280 × 1280 pixels across every experiment. This resolution was selected based on the largest input size of the pre-trained model in YOLOv5 and YOLOv7. It also applied to the DINO algorithm to create a comparable experiment, although the transformer backbone was pre-trained using smaller input images. The model variants of each architecture were considered based on a limited hardware resource on NVIDIA T4 with 16GB GPU memory. A study on the capability of the detector mainly relies on three following models: YOLOv5-l6, YOLOv7-W6, and DINO with Swin-S transformer backbone and Resnet101 backbone. Each object detection approach also provides a variety of preset setups for data augmentation and hyperparameters. However, hyperparameter and augmentation options remain unchanged as default setup or recommendation settings. 4.3 Answer to Q1: Which Is an Appropriate Combination of Preprocessing and Object Detection? This experiment section explored the effect of image preprocessing techniques on the performance of object detection models. We aimed to study whether the preprocessing method is suitable for UAV inspection applications.

(a)

(b)

(c)

Fig. 2. Comparison of image preprocessing methods. a Cropped RGB image focusing on crack area. b Output from Sobel operator. c Output from Canny edge detector.

The experiment attempts to compare the output from the Sobel operator and Canny edge detector on RGB images. The results are shown in Fig. 2, preprocessing methods were performed on raw RGB images at full resolution. Figure 2a illustrates a crack area that was cropped from full resolution image. Preprocessed inputs with the Sobel operator and Canny edge detector are shown in Figure 2b, c, respectively. The output of the Sobel operator shows that unnecessary background details were eliminated while

Automatic Crack Detection Approach for the Offshore Flare …

319

enhancing the metal crack features. In contrast, the result of the Canny edge detector can be clearly noticed that some part of a crack was removed as well. This outcome is caused by hysteresis thresholding which is a part of the Canny edge detector. Therefore, we selected the Sobel operator as a preprocessing step for input data. Table 1. Model performance comparison Model

Params (M)

RGB

Sobel

mAP0.5

mAP0.5:0.95

mAP0.5

mAP0.5:0.95

YOLOv5-m6

35.7

0.6534

0.3132

0.5642

0.2594

YOLOv5-l6

76.8

0.6536

0.321

0.5782

0.278

YOLOv7-W6

70.4

0.654

0.322

0.565

0.2686

DINO-SwinS

69.3

0.592

0.2834

0.478

0.2246

DINO-Resnet101

66.4

0.567

0.2752

0.4516

0.2112

We proceeded to the model training experiment with RGB image and preprocessing method using the Sobel operator. The results are shown in Table 1. The model’s performance was compared using mAP0.5 and mAP0.5:0.95 from COCO metrics. YOLOv5l6 and YOLOv7-W6 achieved approximately 65% mAP0.5 and 32% mAP0.5:0.95 on RGB images. In the transformer-based DINO algorithm, DINO with SwinS backbone has 59.2% mAP0.5 and 28.3% mAP0.5:0.95. Applying a ResNet101 as a backbone gave a slightly lower performance at 56.7% mAP0.5 and 27.5% mAP0.5:0.95. Considering the outcome using the Sobel operator, the performance is drastically dropped when compared to RGB input. The trend of model performance is similar to RGB images in both metrics. YOLOv5 and YOLOv7 still have almost identically to each other and gave higher performance than DINO. Decreasing detection performance is affected by the blurry images that have the consequence of missing edge detection. Due to harsh wind conditions in real operation, a blurry image cannot be avoided in the UAV inspection task. Although an image stabilizer and post-processing are embedded in UAV’s camera, there are some blurry images remained in our data. In summary, the results indicate that the best solution for crack detection is training YOLOv7-W6 using RGB images. YOLOv7-W6 noticeably outperforms the DINO algorithm and has slightly fewer parameters than YOLOv5-l6. Digital image processing techniques like the Sobel operator and Canny edge detector heavily suffered from this scenario. Thus, we recommended performing object detection on raw RGB images.

320

T. Tosawadi et al.

4.4 Answer to Q2: Which Object Detection Approach Is Appropriate for Deployment? In the model deployment phase, memory usage and inference time are critical. These two factors are directly reflected in the budget cost. More memory allocation requires a better GPU for higher memory capacity. The computing power of GPU also correlated with inference time. This section discusses budget efficiency for model deployment on AWS resources. The budget is calculated based on instance type and instance time usage. In our application, a well-trained model will be deployed on ml.g4dn.xlarge AWS EC2 instance which is an entry-level GPU instance. AWS SageMaker Notebooks Instances in the US West (Oregon) region were utilized to conduct all of the experiments in this study. This instance is powered by NVIDIA Tesla T4 with 16 GB GPU memory and price $0.7364 per h. An inferences time is measured during model prediction, not including Non-maximum Suppression (NMS) process. Table 2. Inferences time comparison of each model Method

Number of parameters (M)

Inferences time (ms)

YOLOv5-m6

35.7

68

YOLOv5-l6

76.8

83

YOLOv7-W6

70.4

50

DINO-Swins

69.3

920

DINO-ResNet101

66.4

950

The hardware condition is defined, so less inference time leads to better budget efficiency. Alternatively, sacrificing a model complexity help to reduce inference time but also decreases detection performance. To investigate further on speed-accuracy trade-off, we have trained an additionally available model, YOLOv5-m6. According to inference time in Table 2, the fastest inference time is obtained from YOLOv7-W6 at 50ms. YOLOv7-W6 has 8.35% less parameter than YOLOv5-l6, but inference time is greatly reduced by 39.75%. On the other hand, the DINO algorithm required almost 1 s in both backbones. YOLOv7-W6 is outstandingly faster by 20 times. Comparing the inference time and mAP in Tables 1 and 2, the result shows that YOLOv5-m6 cannot overcome the inference time of YOLOv7-W6 by compensating an accuracy. Although YOLOv5-m6 has 50.71% of YOLOv7-W6 parameters, YOLOv7W6 performed faster in terms of inference time and higher detection performance. From the experiment result, we preferred YOLOv7-W6, which has no trade-off on speed accuracy and gave a faster inference time.

Automatic Crack Detection Approach for the Offshore Flare …

321

4.5 Answer to Q3: Is Image Slicing Approach Is Suitable for Crack Detection on Flare Stack? In drone inspection service, images were taken in wide shots and close shots depending on purposes. A crack on the flare stack is clearly seen in close-shot images. In the wide shot, an image contains the overall detail of inspecting an object but a crack is relatively small to an image scale. Thus, we adopted the slicing inference technique for our crack detection task. The experiment starts by slicing the image in the dataset to 1280 × 1280 pixels with 20% overlapped window size. Then, the models will be trained in sliced images.

(a) full resolution image

(b) sliced image 1

(c) sliced image 2

(d) sliced image 3

Fig. 3. Crack detail comparison between full image (a) and sliced image (b)–(d).

After the slicing process, we found that the crack is unable to be clearly distinguishable. Figure 3b–d are sliced images from the wide shot image shown in Fig. 3a. In Fig. 3d, we ensured that the object is cracked because the sliced images contain the full crack objects. However, we are not capable to identify whether Fig. 3c is a crack or a large metal gap. Training a model using an image with ambiguous features will decrease detection performance. The sliced image in Fig. 3 shows that a slicing inference is not suitable for multi-scale object detection. Especially, a crack detection that requires a full crack defect to ensure correctness. The image slicing also affected the DINO performance where the transformer-based model relied on patches position encoding. 4.6 Compare with the Previous Works All previous research proposed pixel-wise segmentation on concrete, building, and road crack. Their dataset usually contains a clear crack trace that passes through the entire

322

T. Tosawadi et al.

image. Moreover, the crack trace usually stands out from the material surface. In contrast, the sizes of crack traces in our dataset are very diverse, ranging from the small dots to the large crack line strips. The crack traces in our dataset also appear in various patterns, such as dot, strip, rift, and hot tearing, concerning the flare types. All of these patterns are too sensitive for image segmentation.

5 Conclusion This research proposed a combination of input images and object detection models for crack detection on flare stack. The experiments were conducted based on our data collection. The data were gathered from actual operations by the experienced inspector. RGB images were proposed as input along with output from the Sobel operator and Canny edge detector. The inputs were paired with three object detection models, e.g., YOLOv5, YOLOv7, and DINO. The research findings indicate that the Sobel operator and Canny edge detector suffered from a dynamic environment in drone imagery. Among the three input types, RGB images gave a higher outcome in any object detection model. The maximum detection performances are 65.4% mAP0.5 and 32.2% mAP0.5:0.95 which is achievable by training YOLOv7-W6 on RGB images. In terms of inference speed, the fastest inference time at 50 ms was also obtained by using YOLOv7-W6. YOLOv7 has slightly higher mAP and inference faster than YOLOv5. Moreover, YOLOv7 outstandingly outperformed DINO in both mAP and inference speed. Since sliced images are unable to identify whether an object is a crack or metal gap, slicing inference is not recommended in multi-scaled crack detection. From all conducted experiments, YOLOv7 with raw RGB image is an appropriate approach that satisfied detection performance and inference speed. Acknowledgments. This project was supported by PTT Exploration and Production Public Company Limited (PTTEP) and AI and Robotics Ventures Co., Ltd. (ARV). All in all, We would like to express our gratitude to Skyller Solutions Company for providing us with the dataset.

References 1. Zhang, H. et al.: DINO: DETR with improved DeNoising anchor boxes for end-to-end object detection. arXiv: 2203.03605 [cs.CV] (2022) 2. Liu, S., et al.: DAB-DETR: dynamic anchor boxes are better queries for DETR. In: International Conference on Learning Representations. https://openreview.net/forum?id=oMI9Pj Ob9Jl (2022) 3. Li, F., et al.: DN-DETR: accelerate DETR training by introducing query denoising. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 13619–13627 (2022) 4. He, K., et al.: Deep residual learning for image recognition. arXiv (2015). https://doi.org/10. 48550/ARXIV.1512.03385 5. Liu, Z., et al.: An image is worth 16 × 16 words: transformers for image recognition at scale. CoRR (2021). https://doi.org/10.48550/arXiv2010.11929

Automatic Crack Detection Approach for the Offshore Flare …

323

6. Redmon, J., et al.: You only look once: unified, real-time object detection. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016). https://doi.org/10. 1109/cvpr.2016.91 7. Redmon, J., Farhadi, A.: YOLO9000: better, faster, stronger. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6517–6525 (2017). https://doi.org/ 10.1109/CVPR.2017.690 8. Redmon, J., Farhadi, A.: YOLOv3: an incremental improvement arxiv.org/abs/1804.02767 (2018) 9. Bochkovskiy, A., Wang, C.-Y., Liao, H.-Y.: YOLOv4: optimal speed and accuracy of object detection (2020) 10. Ultralytics/yolov5, Oct 2020. https://doi.org/10.5281/zenodo.4154370 11. Li, C., et al.: YOLOv6 v3.0: a full-scale reloading. https://doi.org/10.48550/ARXIV/2301. 05586. arxiv.org/abs/2301.05586 (2023) 12. Wang, C.-Y., Bochkovskiy, A., Mark Liao, H.-Y.: YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. ArXiv preprint arXiv:2207.02696 (2022) 13. Munawar, H.S., et al.: Civil infrastructure damage and corrosion detection: an application of machine learning. Buildings 12. https://doi.org/10.3390/buildings12020156 14. Zou, Q., et al.: DeepCrack: learning hierarchical convolutional features for crack detection. IEEE Trans. Image Process. 1. https://doi.org/10.1109/TIP.2018.2878966 15. Huyan, J., et al.: CrackU-net: a novel deep convolutional neural network for pixelwise pavement crack detection. Struct Control Health Monit 27, e2551. https://doi.org/10.1002/stc. 2551

The Recent Trend of Artificial Neural Network in the Field of Civil Engineering Aditya Singh(B) School of Civil Engineering, Lovely Professional University, Phagwara, India [email protected]

Abstract. This paper explains the concept of Artificial Neural Network briefly in general and it also particularly focuses the current trends and applications of ANN in the area of civil engineering for various purposes. The author also performed a comprehensive review of numerous scientific and research papers published in the recent years relevant to the topic of the study, in order to find the gaps in the research. Further, the author collected data from various sources in order to perform graphical analysis and support the study. Major benefits and challenges of using Artificial Neural Network are analysed on the basis of the graphs shown in the paper. Keywords: Artificial neural network · Convolutional neural network · Civil engineering · Recurrent neural network · Structural engineering

1 Introduction Artificial Neural Network is a term which was motivated from a biological human brain and it is a subfield of Artificial Intelligence. Generally it is a computational network which was made on the basis of biological neural networks which are responsible in making the brain’s structure. The way human brain has neurons which are interconnected to the other neurons, the same way Artificial Neural Networks also have neurons which are connected to the other neurons in different layers of the networks, such neurons present in Artificial Neural Networks are called as Nodes [24]. According to the 1st neurocomputer’s inventor Dr. Robert Hecht-Nielsen, a neural network can be described as a computing system which was created by various simple as well as highly interconnected processing components that has the ability to process info with the help of their vibrant state response to the external inputs received [16]. 1.1 Artificial Neural Network’s Basic Structure Similar to the way the brain of a human works, Artificial Neural Networks was an artificial copy of it with the assistance wires as well as silicon which acted as living dendrites as well as neurons. In case of the brain of a human, it is made of around eighty six billion nerve cells which are known as neurons. These are then linked to © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 324–333, 2024. https://doi.org/10.1007/978-3-031-50158-6_32

The Recent Trend of Artificial Neural Network

325

other thousand cells which are called as Axons. Dendrites accepts the stimuli from an external environ or simply said inputs with the help if sensory organs. Then, the earlier mentioned inputs are able to generate electric impulses that can rapidly travel through the given neural network. Then it depends upon the neuron to choose whether to not send the message forward or to send it forward for the next neuron to handle it. Then, in the case of Artificial Neural Networks, they are made of multiple nodes that has the ability to copy human brain’s biological neurons. These neurons with the help of links are connected and it allows the neurons to interact with the other neurons smoothly. Then the nodes are able to receive the input data, in order to execute simple operations on it. Then such operation’s results are further passed to the next neurons. Node Value or Activation is named to the output generated at each such nodes. It is said that each link is connected with Weight. Artificial Neural Networks have the ability to learn which happens by changing the values of weights [16]. 1.2 Is There Any Need to Use Advanced Technologies like ANN in the Area of Civil Engineering? Advanced technologies like Artificial Neural Network is able to help in solving complex problems and hidden problems which civil engineers normally face on a regular basis. For instance, Fuzzy logic, ANN are one of the important parts of ITS, which is also a major advancement in traffic and transportation engineering, which is a major branch of civil engineering [11, 12]. Since the world is going in the direction of Construction 4.0, where Digitalization of the construction sector is important, which is also a subfield of civil engineering [13]. So, ANN will also contribute towards the digital transformation of the construction sector. These technologies will not only contribute towards the development of the areas of civil engineering, but also improve the efficiency, quality of work, etc. in the civil engineering domain. 1.3 Objectives In this paper, the following objectives are considered: • To understand the concept of Artificial Neural Network and its types. • To understand the market trends and value of Artificial Neural Network not only in present but also in the future. • To understand the various subfields of Civil Engineering where Artificial Neural Networks can be applied.

2 Motivation The author had observed that Artificial Neural Network is becoming more and more popular as the time passes as well as the market of Artificial Neural Network is expected to grow rapidly in the Asia Pacific region in the future. Then, a lot of research is still being going on to apply Artificial Neural Network in the subfields of Civil Engineering. Being a Civil Engineer, the author felt that it is necessary to highlight and understand the concept of ANN as well as its growing applicability in area of Civil Engineering along with expected increase in the market value in the future.

326

A. Singh

3 Literature Review De Souza et al. worked on comparing adaptive as well as non-adaptive optimization algorithms aimed at ANN training which can be used for the purpose of damage diagnosis especially in the civil structures [1]. Yang et al. studied ANN and they worked on a review paper where ANN was applied in a subfield of civil engineering which was transportation engineering, particularly in pavement engineering area [2]. Falcone et al. studied ANN and they worked in regards to the technical feasibility forecast of seismic retrofitting in the already present Reinforced Concrete Structures [3]. Freitag et al. studied ANN and they utilized it on optimizing structural topologies which was reliability based one [4]. Wang et al. studied ANN and they amalgamated it with damage parameters in order to forecast beginning lifetime of fretting fatigue cracks [5]. Safoklov et al. studied ANN and they used it on overhaul in addition on aircraft maintenance repair model [6]. Li et al. studied ANN and they used it to forecast the thermal conductivity of the given soils which was done on the basis of an orderly database [7]. Khaleghi et al. studied ANN and they used it to characterize fracture from the given noisy displacement data [8]. El Jery et al. studied ANN and they used it with numerical simulation to forecast heat transfer as well as hydrodynamic in a given geothermal heat exchanger in order to attain a certain tube’s optimal diameter with the minimum entropy by utilizing water nanofluid or Al2 O3 as well as water [9]. Jin et al. studied ANN and they used it to forecast the chloride diffusivity of a certain aggregate concrete which is a recycled one [10].

4 Research Gaps Based on the recent advancement of the technology many researchers are implementing ANN successfully in the various branches of civil engineering, which can be observed through the recent scientific and research publications. However, particularly in the branches of civil engineering like water resource engineering as well as construction technology and management, the current researches are insufficient and more research can be done using ANN in the future.

5 Main Focus of the Paper Along with Issues and Problems The author explained the concept of Artificial Neural Network and its basic structure in this paper. The author covered the history of ANN briefly and major type of ANN in the paper. Current market trend of ANN was highlighted by performing graphical analysis by the author. Some major benefits of using ANN in the area of civil engineering along with major problems were discussed. However, it can’t be ignored that many civil engineers are still unfamiliar about ANN or the ways to implement it practically in a project. Then terrain and location problems along with long time required to learn ANN for engineers with different background and so on, creates major obstacles in applying ANN on a large scale in civil engineering projects.

The Recent Trend of Artificial Neural Network

327

6 Different Types of Artificial Neural Networks There are various kinds of artificial neural networks, which are as follows [17]: • • • • •

Recurrent Neural Networks Modular Neural Networks Convolutional Neural Networks Feed-forward Neural Networks De-convolutional Neural Networks.

Recurrent Neural Networks They are a complex type of neural network, as they work by saving the processing node’s output as well as feed the acquired result back into the given model. Through this way the mode learns the way to forecast a given layer’s outcome. Every node in this model behaves as a memory cell, which continues the computation in addition to execution of the given operations. Modular Neural Networks They consists of numerous neural networks which work independently, without relying on the other neural networks. At the time of the computation process, these networks never inhibit the activities of the other networks or they don’t even communicate with the other networks. Convolutional Neural Networks They are considered as part of the most prevalent models which are actively used at present. They uses a variation of multilayer perceptron in addition to it consists of at least one or more than one convolutional layers which has the ability to completely pooled or connected. Feed-Forward Neural Networks They are the simplest type of neural networks. With the help of numerous input nodes, they pass info in one given direction till the point it reaches the output node. FFNN can have or hidden node layers or they might not even have it, which also makes its functioning highly interpretable. De-convolutional Neural Networks They can be said opposite of CNN model as they use a reverse of the latter model process. Their goal is to search for the missing signals or features which initially might have been thought as insignificant to the system’s task of CNN.

7 Methodology In this section the recent trends and applications of Artificial Neural Network applied in the various branches of civil engineering is explained through the following flow chart shown in Fig. 1.

328

A. Singh Artificial Neural Network

Civil Engineering

Water Resource Engineering: detecting loss of water through cracks (pipe system present in underground); analysis, modelling as well as mathematical formulation of water distribution systems; rehabilitation and retrofitting of water networks.

Geotechnical Engineering: examination of a) liquefaction of soil, b) foundation studies, c) soil – pile interactions, d) loading conditions as well as potential failures.

Disaster Management: prediction of natural calamities.

Construction Technology and Management: project bidding, quality inspection systems, project planning and management, analysis of welding quality, appraisal of real estate, etc.

Structural Engineering: studying of concrete behaviour: shear strength development based on beam’s different dimensions, crack propagation, failure prediction. Structural health monitoring, composite structure modelling and so on.

Transportation Engineering: control, automobile scheduling, truck brake system diagnosis, routing systems, automotive, crack propagation on concrete pavement, road safety and accidents.

Fig. 1. Showing the trends and applications of ANN in the area of civil engineering [14, 16]

8 Result and Discussion In this section, data from various sources like Data Bridge, Emergen Research, etc. are collected by the author to perform graphical analysis, in order to support the study (Tables 1, 2 and 3).

9 Advantages of Artificial Neural Network In this section, some major advantages of Artificial Neural Network particularly in the area of civil engineering are discussed, which are as follows: • According to Fig. 2, the market size of Artificial Neural Network from the year 2019 is sharply increasing till the year 2030, which means that more and more application of ANN is required in the various branches of civil engineering which might definitely be helpful in the growth of civil engineering.

The Recent Trend of Artificial Neural Network

329

Table 1. Showing the comparison of expected market size of ANN by different sources [15, 18–22, 25] Market size of ANN

In million USD

Year 2019 by MarketsandMarkets

117

Year 2021 by Emergen Research

160.8

Year 2021 by Proficient Market Insights

169.3

Year 2021 by Data Bridge

171.58

Year 2021 by Maximize Market Research Pvt. Ltd.

177.49

Year 2022 (estimated) by Research and Markets

171.88

Year 2024 (expected) by MarketsandMarkets

296

Year 2027 (expected) by Research and Markets

258.45

Year 2028 (expected) by Proficient Market Insights

542.3

Year 2029 (expected) by Maximize Market Research Pvt. Ltd.

763.2

Year 2029 (expected) by Data Bridge

793.69

Year 2030 (expected) by Emergen Research

743

Table 2. Showing the expected and past NN software market [23] Neural network software market Year 2020 Year 2026 (expected)

In (USD) 8,300,000,000 50,660,000,000

Table 3. Comparing market revenue share of North America with the rest of the world [18] Market revenue share

In (%)

North America in 2021

30.8

Rest of the world in 2021

69.2

• According to Fig. 3, the software market of Neural Network is expected to grow tremendously by the year 2026, which means that more NN software must be utilised in the various branches of civil engineering to improve the efficiency and quality aspects. • According to Fig. 4, the market revenue share of ANN in North American continent is significantly high in the world, which can also be related to the advancement and development of the North American continent is also higher than most of the countries present in other continents of the world.

330

A. Singh

Market Size of ANN 800 600 400 200 0 In Million USD

Fig. 2. Showing the comparison of expected market size of ANN by different sources [15, 18–22, 25]

Neural Network Soware Market 50,660,000,000 60,000,000,000 40,000,000,000 20,000,000,000

8,300,000,000

0 In (USD) Year 2020

Year 2026 (Expected)

Fig. 3. Showing the expected and past NN software market [23]

• According to Figs. 2 and 3, there is a major expansion of ANN through its growing practical applications or the need and use of various types of NN software for different situations, similar growth of ANN can be expected in the area of civil engineering in the future. • Based on the above figures, positive trend of ANN can be seen in the future, so it will be wise to focus on the practical development and implementation of ANN in the area of civil engineering as well.

The Recent Trend of Artificial Neural Network

331

Market Revenue Share 69.2 In (%)

30.8

0

10

20

30

Rest of the world in 2021

40

50

60

70

North America in 2021

Fig. 4. Comparing market revenue share of North America with the rest of the world [18]

10 Major Challenges of ANN in the Area of Civil Engineering In this section, some major challenges of using ANN in the civil engineering area, are: • Based on Figs. 2 and 3, the traditional civil engineers might face difficulty in getting used to ANN in their day to day work in the future. • According to the above figures, civil engineers are needed to learn ANN in their university courses or some practical training is required for them to successfully implement ANN in their domain based on different situations in the future. • Incorporating ANN on a large scale civil engineering works and projects will also be a big challenge for people who are familiar to accomplishing their tasks though traditional ways.

11 Conclusion In this paper, the basic concept of ANN was explained and its structure was discussed. The author talked about the types of ANN in brief in the paper. The trends and applications of ANN in the subfields of Civil Engineering was briefly mentioned in this paper. Further, the author performed some graphical analysis after collecting data from different sources to check the future trends, market value of ANN in the paper. Then, some major advantages as well as challenges of using ANN in the area of civil engineering were discussed in short in the paper, based on the analysis of the graphs. As ANN is expected to grow tremendously in the future, it is better to make all the civil engineers prepared to adapt and learn ANN technologies in order to implement them in projects at will.

332

A. Singh

References 1. de Souza, C.P.G., Kurka, P.R.G., Lins, R.G., de Junior, J.M.: Performance comparison of nonadaptive and adaptive optimization algorithms for artificial neural network training applied to damage diagnosis in civil structures. Appl. Soft Comput. 104, 107254 (2021) 2. Yang, X., Guan, J., Ding, L., You, Z., Lee, V.C.S., Mohd Hasan, M.R., Cheng, X.: Research and applications of artificial neural network in pavement engineering: a state-of-the-art review. J. Traffic Transp. Eng. (Engl. Ed.) 8(6), 1000–1021 (2021) 3. Falcone, R., Ciaramella, A., Carrabs, F., Strisciuglio, N., Martinelli, E.: Artificial neural network for technical feasibility prediction of seismic retrofitting in existing RC structures. Structures 41, 1220–1234 (2022) 4. Freitag, S., Peters, S., Edler, P., Meschke, G.: Reliability-based optimization of structural topologies using artificial neural networks. Probab. Eng. Mech. 70, 103356 (2022) 5. Wang, C., Li, Y., Tran, N.H., Wang, D., Khatir, S., Wahab, M.A.: Artificial neural network combined with damage parameters to predict fretting fatigue crack initiation lifetime. Tribol. Int. 175, 107854 (2022) 6. Safoklov, B., Prokopenko, D., Deniskin, Y., Kostyshak, M.: Model of aircraft maintenance repair and overhaul using artificial neural networks. Transp. Res. Proc. 63, 1534–1543 (2022) 7. Li, K.-Q., Kang, Q., Nie, J., Huang, X.: Artificial neural network for predicting the thermal conductivity of soils based on a systematic database. Geothermics 103, 102416 (2022) 8. Khaleghi, M., Haghighat, E., Vahab, M., Shahbodagh, B., Khalili, N.: Fracture characterization from noisy displacement data using artificial neural networks. Eng. Fract. Mech. 271, 108649 (2022) 9. El Jery, A., Khudhair, A.K., Abbas, S.Q., Abed, A.M., Khedher, K.M.: Numerical simulation and artificial neural network prediction of hydrodynamic and heat transfer in a geothermal heat exchanger to obtain the optimal diameter of tubes with the lowest entropy using water and Al2 O3 /water nanofluid. Geothermics 107, 102605 (2023) 10. Jin, L., et al.: Prediction of the chloride diffusivity of recycled aggregate concrete using artificial neural network. Mater. Today Commun. 32, 104137 (2022) 11. Bhowmik, S., Singh, A., Misengo, C.: A case study on intelligent transport system using traffic lights. Our Herit. 67(7), 96–110 (2019) 12. Singh, A.: Importance of fuzzy logic in traffic and transportation engineering. In: Vasant, P., Zelinka, I., Weber, G.W. (eds.) Intelligent Computing and Optimization. ICO 2021. Lecture Notes in Networks and Systems, vol. 371, pp. 87–96. Springer, Cham (2022) 13. Singh, A.: The significance of digitalization of the construction sector. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing and Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569, pp. 1069–1077. Springer, Cham (2022) 14. CivilDigital: https://www.google.com/amp/s/civildigital.com/all-about-artificial-neural-net work-ann-in-civil-engineering/amp/ 15. Data Bridge Market Research: https://www.google.com/amp/s/www.databridgemarketresea rch.com/reports/global-artificial-neural-network-ann-market/amp 16. Tutorialspoint: https://www.tutorialspoint.com/artificial_intelligence/artificial_intelligence_ neural_networks.htm 17. TechTarget: https://www.google.com/amp/s/www.techtarget.com/searchenterpriseai/defini tion/neural-network%3famp=1 18. Emergen Research: https://www.emergenresearch.com/amp/industry-report/artificial-neuralnetwork-market 19. MarketsandMarkets: https://www.marketsandmarkets.com/Market-Reports/artificial-neuralnetwork-market-21937475.html

The Recent Trend of Artificial Neural Network

333

20. BioSpace: https://www.biospace.com/article/artificial-neural-network-market-size-worthusd-743-0-million-in-2030-emergen-research-/ 21. GlobeNewsWire: https://www.globenewswire.com/news-release/2022/11/14/2554650/0/en/ Artificial-Neural-Networks-Market-2022-Will-Revenue-to-Cross-reach-US-542-3-millionby-2028-Research-by-Business-Opportunities-Top-Companies-report-covers-Market-spe cific-challenge.html 22. Business Wire: https://www.businesswire.com/news/home/20220920006083/en/Global-Art ificial-Neural-Network-Market-to-Reach-258.45-Million-by-2027---ResearchAndMarkets. com 23. Mordor Intelligence: https://www.mordorintelligence.com/industry-reports/neural-networksoftware-market 24. Javatpoint: https://www.javatpoint.com/artificial-neural-network 25. Maximize Market Research: https://www.maximizemarketresearch.com/market-report/glo bal-artificial-neural-network-market/83873/

Analyzing Price Forecasting of Grocery Products in Bangladesh: A Comparison of Time Series Modeling Approaches Md Mahmudul Hoque1 , Ikbal Ahmed1 , Nayan Banik2 and Mohammed Moshiul Hoque3(B)

,

1 Department of Computer Science and Engineering, CCN University of Science and

Technology, Kotbari, Cumilla 3506, Bangladesh 2 Department of Computer Science and Engineering, Comilla University, Kotbari,

Cumilla 3506, Bangladesh 3 Department of Computer Science and Engineering, Chittagong University of Engineering and

Technology, Chittagong 4349, Bangladesh [email protected]

Abstract. Bangladesh is a south Asian agriculturally reliant country producing essential grocery products. Among such products, the most notable commodities are rice, wheat flour, and lentils (massor). The unforeseen increase in those product prices puts wholesalers, retailers, dealers, and buyers in tension as the purchasing ability remains the same. Several factors are liable in this scenario, making price forecasting by category an extremely challenging task for decision-makers. Investigation of underlying factors helps policymakers monitor product prices and keep the market in equilibrium. Several studies addressed this issue from Bangladesh’s perspective. However, there is still scoped to perform further analysis to better model the forecasting premises. This study focuses on using multiple time series analysis of three common grocery goods (rice, lentils, and wheat flour) and comparing several modeling approaches by Mean Absolute Error (MAE). The experiment covers forecasting of items in divisional regions on a monthly basis. Evaluation of the results showed that the RNN and Naïve Ensemble model accurately generalizes price changes. This work aims to help future research in this domain by conveying guidelines and comparative analysis of other modern approaches. Keywords: Price forecasting · Prediction · Machine learning · Time series analysis · Evaluation

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 334–341, 2024. https://doi.org/10.1007/978-3-031-50158-6_33

Analyzing Price Forecasting of Grocery Products …

335

1 Introduction Though most people in Bangladesh live within the food purchase limit, 14% of the people are still living in poverty [6]. Many low-income individuals, particularly those who live on the borderline, are critical sufferers. The general people have been protesting the uncontrolled increase in commodity prices. Nonetheless, the price fluctuation of grocery items has increased substantially due to alliances. The general population has recently been spotted lining up behind Trading Corporation of Bangladesh (TCB) cars in search of fair-priced goods. The most talked-about subject right now is the unexpected spike in grocery costs. Therefore, forecasting grocery items is essential nowadays. In addition, predicting for businesses, the weather, risk management, national development, and grocery prices are commonplace today. There is a total area of 15 million hectares, more than half of Bangladesh’s land used for agricultural purposes [12]. Bangladesh, a selfsufficient country in food production, heavily relies on a few staple commodities such as rice, wheat flour, and lentils, which are an integral part of the daily diet. However, due to its developing and middle-income status, many farmers and low-income individuals struggle to afford these essential commodities when prices are high. Although the government is trying to curb the price hike caused by unscrupulous traders, it is not always successful. Therefore, time series forecasting is essential at all industry levels. This study analyzes using machine learning techniques to forecast commodity prices based on the purchase history. There are numerous time series forecasting algorithms, including Facebook Prophet (FB Prophet), Long Short-Term Memory (LSTM), Exponential Smoothing, Autoregressive Integrated Moving Average (Arima), and Seasonal Autoregressive Integrated Moving Average (Sarima) [2]. In this study, we used 12 Time Series algorithms and Compared them by measuring their performance (MAE). Predicting the prices of these essential commodities in advance could greatly aid in regulating the market prices, as the general public would be willing to purchase them at the predetermined price, and sellers would be obliged to sell at that price. Such an approach could potentially disrupt market syndicates that aim to manipulate prices to their advantage. This study will benefit individuals who plan to work in this field in the future and those who set the retail price of a product.

2 Related Work Hasan et al. [8] focused on predicting rice prices based on one-year data. They investigated five machine learning methods (KNN, NB, DT, SVM, and RF), where RF achieved the highest prediction accuracy (70%). Hossain et al. [9] used CNN and linear regression models to predict paddy prices. They surveyed local markets for predictions of paddy prices and found some factors for fluctuations in paddy prices. Ganesan et al. [1] proposed a model to make a day-ahead prediction of food sales in one of the top multiplexes in India. They investigated prices using XGBoost, Artificial Neural Network (ANN) Extra Trees, and LSTM. To predict airfare prices in Greece, Tziridis et al. [13] applied eight machine learning algorithms to predict air ticket prices where the Bagging regression tree performed best than other techniques by gaining 87.42% accuracy. Pudaruth et al. [11] employed various ML techniques to predict the price of used cars in Mauritius.

336

M. M. Hoque et al.

They employed four machine learning techniques (LR, KNN, Naïve Bayes, and Decision Trees) to anticipate used car prices. Didem et al. [7] used LSTM and Facebook Prophet (FBP) algorithm to forecast future trends in Brent oil prices based on historical prices to improve accuracy and stability. The authors gathered weekly Brent crude oil price data from the NASDAQ Commodities. The LSTM model performed well in accuracy in predicting brent oil, with an R2 value of 0.89. Chadalavada et al. [3] predicted electricity demand for households, offices, or buildings using the FBP and Autoregressive Integrated Moving Average (Arima) model. The data was gathered from the UCI Machine Learning Repository, and they predicted data for one month. The accuracy of different models varies depending on the dataset.

3 Methodology Figure 1 illustrates the proposed approach of price prediction consisting of data collection, preprocessing, and model training.

Fig. 1. Proposed method of price prediction

3.1 Data Collection The information was gathered from the Humanitarian Data Exchange (HDX) website [4]. It is a free online platform that allows users to share data from various disasters and organizations [5]. There are 18,735 datasets available on the HDX websites. However, we will be working with Bangladeshi food prices. The HDX collects information from the “World Food Programme (WFP)” and updates the dataset monthly based on WFP Market Monitoring data. WFP is a UN agency that monitors domestic food prices in more than 80 countries [10]. Every month, the WFP publishes a “Market Monitor” report; this organization works in several areas, including market monitoring. The United Nations World Food Programme oversees food aid. It is the world’s largest humanitarian organization dedicated to preventing hunger and ensuring food security [14]. Rice, wheat flour, lentils, and palm oil include 7732 entries and 14 wholesale and retail pricing attributes. Table 1 shows the statistics of the dataset.

Analyzing Price Forecasting of Grocery Products …

337

Table 1. Dataset statistics Location

Period

Total samples

Dhaka

Nov 2005–Dec 2021

2584

Chattogram

Jan 2006–Dec 2021

1110

Rajshahi

Jul 2006–Dec 2021

1040

Khulna

Oct 2005–Dec 2021

1063

Barishal

Dec 2006–Dec 2021

825

Sylhet

Jan 2007–Dec 2021

763

3.2 Data Preprocessing For data preparation, we utilized the Scaler transformer in Darts. Before converting it to a time series object in Darts, the scaler first constructed a random time series with 100 data points. Then use the ‘fit’ technique to fit an instance of the Scaler transformer to the time series. The transform technique transforms the time series to the desired range. The parameters determined during the relevant phase apply the scaling transformation to the time series. The ‘inverse transform’ technique returned the scaled time series to its original scale. Table 2 illustrates various features and their corresponding dataset descriptions. However, it should be noted that additional attributes within the dataset have not been included in this table. These attributes include Unit, Currency, Country, Unit ID, Market ID, Commodity ID, and Category ID. While these attributes may not have been explicitly discussed, they are nevertheless essential components of the dataset and should be considered when analyzing the data. 3.3 Model Training This study exploited 12 algorithms, including TCN (Temporal Convolutional Network), NBEATS (Neural Basis Expansion Analysis for Interpretable Time Series Forecasting), (T) BATS (Trend, Seasonal, and Exogenous (external) Bayesian Time Series), AutoARIMA, Theta (Thresholding Expectation-Maximization), Naïve Drift, Naïve Seasonal, Facebook Prophet, ARIMA, FFT (Fast Fourier Transform) RNN (Recurrent Neural Network). The suggested work modified several factors to reduce forecasting inaccuracy. For instance, 40 epochs were sufficient to minimize error for NBEATS when using Temporal Convolutional Network, where epochs were 400. RNN required 500 epochs to perform better, in any case.

338

M. M. Hoque et al. Table 2. Attribute and feature description of the dataset

Attribute name Description date

The primary attribute within this dataset is a monthly time series data

cmname

The name of commodities is a important attribute within this dataset, which includes four distinct items: rice, wheat flour, Lentils, and Palm oil. The attribute includes pricing information for both wholesale and retail categories for each commodity

category

The dataset comprises four commodity items, classified into three categories: ‘cereals and tubers’, ‘pulses and nuts’, and ‘oil and fats’

admname

The name of regions is another attribute within this dataset, which includes seven primary regions. However, for this study, we utilized data from six of these regions

price

The price of commodities is the target attribute within this dataset

mktname

The market name is another crucial attribute within this dataset, which includes divisional and district-wise information. However, for the purposes of this research, we focused solely on the divisional data

4 Results and Discussion To implement the time series algorithms, ‘Darts’ was employed, while graphical representations were visualized using Plotly Express. 70% of the data were utilized for training, while the remaining 30% were used for validation. This study uses the MAE (Mean Absolute Error) measure to evaluate the performance. Figure 2 shows the percentage of the MAE score of rice price prediction. Results show that the MAE score for Sylhet is 1.92, whereas it is 2.86 for Dhaka. In contrast, Chattogram, Rajshahi, Khulna, and Barishal divisions of the RNN model behaved well, with corresponding MAE scores of 1.99, 2.79, 2.47, and 1.98. The Naïve Ensemble model performed admirably and provided reduced inaccuracy for the divisions of Sylhet and Dhaka. Figure 3 shows the prediction of wheat flour in various divisions. For the Chattogram and Rajshahi Divisions, the RNN model performed as expected better than any other model, with MAE scores of 1.37 and 0.59, respectively. On the other hand, Naïve Ensemble provided the Khulna and Barishal divisions with less inaccuracy, and the MAE scores are 0.62 and 0.71, respectively. Out of 12 models, the TCN and NBEATS model impressively outperformed them all for the Dhaka and Sylhet divisions. The MAE score for the Sylhet division of the NBEATS model is 0.38, while the TCN model’s score for the Dhaka division was 0.79.

Analyzing Price Forecasting of Grocery Products …

339

Fig. 2. Prediction of rice prices (MAE score)

Fig. 3. Prediction of wheat flour prices (MAE score)

Figure 4 shows the percentage of the MAE score each model for lentils received in each division. The chart compares the MAE scores for models with a 12-times series. The chart shows that the Naïve Ensemble model performed exceptionally well across all divisions and that RNN also provided reduced error while achieving the same score as the Naïve Ensemble for the Barishal division. The Barishal division of RNN and Naïve Ensemble received a 4.97 MAE score. Moreover, Naïve Ensemble received MAE

340

M. M. Hoque et al.

scores of 4.83 in the Dhaka division, 4.54 in the Chattogram division, 5.03 in the Rajshahi division, and 4.99 in the Khulna 5.65 in the Sylhet division.

Fig. 4. Prediction of lentils prices (MAE score)

5 Conclusion Commodity prices are unpredictable and usually increasing in nature in Bangladesh. Several factors (such as increased demand, lower supply, rising production costs, unfair warehousing practices, and lacking market monitoring) contribute to grocery items’ price increases. Moreover, the prices of grocery items varied from time to time. This work investigated the prices of the three most consumed items (rice, wheat flour, and lentils) in six divisions of Bangladesh. Multiple time series regressor models are exploited to analyze the monthly periodic data and evaluate each model’s performance using MAE. Results showed that the RNN and Naïve ensemble models performed better in predicting prices in most cases. In the future, we aim to utilize deep learning techniques to forecast more grocery items weekly with greater accuracy.

Analyzing Price Forecasting of Grocery Products …

341

References 1. Adithya Ganesan, V., Divi, S., Moudhgalya, N.B., Sriharsha, U., Vijayaraghavan, V.: Forecasting food sales in a multiplex using dynamic artificial neural networks. In: Science and Information Conference, pp. 69–80. Springer (2019) 2. Advancing Analytics: 10 incredibly useful time series forecasting algorithms, 15 Feb 2022. https://www.advancinganalytics.co.uk/blog/2021/06/22/10-incredibly-useful-time-ser ies-forecasting-algorithms 3. Chadalavada, R., Raghavendra, S., Rekha, V.: Electricity requirement prediction using time series and Facebook’s prophet. Indian J. Sci. Technol. 13(47), 4631–4645 (2020) 4. Humanitarian Data Exchange: Bangladesh—food prices. https://data.humdata.org/dataset/ wfp-food-prices-for-bangladesh 5. Humanitarian Data Exchange: The humanitarian data exchange (2014). https://data.humdata. org/ 6. Giménez, L., Jolliffe, D., Sharif, I.: Bangladesh, a middle income country by 2021: what will it take in terms of poverty reduction? Bangladesh Dev. Stud. 37(1 & 2), 1–19 (2014) 7. Güleryüz, D., Özden, E.: The prediction of brent crude oil trend using lstm and facebook prophet. Avrupa Bilim ve Teknoloji Dergisi 20, 1–9 (2020) 8. Hasan, M.M., Zahara, M.T., Sykot, M.M., Nur, A.U., Saifuzzaman, M., Hafiz, R.: Ascertaining the fluctuation of rice price in Bangladesh using machine learning approach. In: 2020 11th International Conference on Computing, Communication and Networking Technologies (ICCCNT), pp. 1–5. IEEE (2020) 9. Hossain, K.A., Raihan, M., Biswas, A., Sarkar, J.P., Sultana, S., Sana, K., Sarder, K., Majumder, N.: Paddy disease prediction using convolutional neural network. In: Intelligent Computing and Optimization: Proceedings of the 4th International Conference on Intelligent Computing and Optimization 2021 (ICO2021), vol. 3, pp. 268–276. Springer (2022) 10. World Food Programme: Where we work (1961). https://www.wfp.org/countries 11. Pudaruth, S.: Predicting the price of used cars using machine learning techniques. Int. J. Inf. Comput. Technol. 4(7), 753–764 (2014) 12. Sarkar, J.P., Raihan, M., Biswas, A., Hossain, K.A., Sarder, K., Majumder, N., Sultana, S., Sana, K.: Paddy price prediction in the South-Western region of Bangladesh. In: Intelligent Computing and Optimization: Proceedings of the 4th International Conference on Intelligent Computing and Optimization 2021 (ICO2021), vol. 3, pp. 258–267. Springer (2022) 13. Tziridis, K., Kalampokas, T., Papakostas, G.A., Diamantaras, K.I.: Airfare prices prediction using machine learning techniques. In: 2017 25th European Signal Processing Conference (EUSIPCO), pp. 1036–1039. IEEE (2017) 14. (WFP): Un world food programme (1961). http://wfp.org

Big Data: Identification of Critical Success Factors Leo Willyanto Santoso(B) Petra Christian University, 121–131 Siwalankerto, Surabaya 60296, Indonesia [email protected]

Abstract. Nowadays, the organization try to implement big data technology. By implementing it, they can develop big data analytics as a tools for competitive advantage. Top management realize this phenomenon and should respond to the challenges of information revolution. Data-driven business decision gains more and more prominence. Data could be used to see exactly what’s happening in organization and use the information to make the business more agile. However, there are several factors that are critical for a Big Data project’s success. The objective of this research is to recognise the important critical success factors to obtain organizational value from big data. We evaluate each factor from panel that consist of 18 professionals to make a rank based on the degree of importance by using Delphi method. Delphi methodology was divided in three phases and two rounds. The test result revealed that there are five main critical success factors are: (1) aligning big data strategy to customer insights and business goals; (2) having adequately IT infrastructure; (3) having technologies and tools used to enable big data analytics processes; (4) data-driven culture in the organizations and management; (5) automated or semi-automated data analytics tool. Keywords: Big data · Critical success factors · Business goals · Organizational value · Delphi

1 Introduction New technologies capable of collecting and analyzing data know and understand the client better, leading to a decision-making process more informed. According to IBM, around of 2.5 quintillion bytes of data every day [1]. The same source and consultant Gartner further points out that by 2015 4.4 million data scientists will be needed [2]. It can be said that the importance of optimizing management of data within the organization, becoming a critical factor in success in obtaining competitive advantage [3]. The combination of the Internet and democratization of the creation and different formats, has given rise to many types of data. The data thus increased in volume, variety, and velocity. This phenomenon has been called the big data [4]. The potential impact of big data on organizations continues to increase, the management structures and processes, thus giving rise to the need for basic and applied research among the various disciplines (such as information systems, marketing, accounting, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 342–349, 2024. https://doi.org/10.1007/978-3-031-50158-6_34

Big Data: Identification of Critical Success Factors

343

human Resources). It will be necessary develop and improve analytical methods appropriate to big data consider the volume, variety, speed and accuracy of data. Although many organizations can use the provided insights [5], it is also required to consider the increasing importance of data to improve the supply of products and services [6]. Thus, the research question of this study is “what are the critical success factors more important to withdraw value of the big data?”. Intends clarify the factors that benefit a big data strategy as well as organizational benefits at a time when unstructured data grows 15 times more than structured data [7]. When using the Delphi methodology, this research intends to elaborate an exploratory study [4, 8]. The study has the following structure: first, a review of the literature on big data in Sect. 2, a brief critic of success, critical success factors of IS, and finally a focus on the critical success factors for withdrawing big data value. Section 3 discusses the methodology, the presentation of the panel, as well as collection and processing of data. Section 4 presents the analysis of the results obtained, discussion and confrontation with the literature. Finally, Sect. 5 expresses the conclusions, practical implications, limitations and research related to the subject under study.

2 Critical Success Factors in IS and Big Data The terminology of critical success factors has acquired different interpretations since it became popular in the 1980s [9]. This author defined them as areas whose results, if they are satisfactory, allow a competitive performance of the organization. In this way, it is essential to identify the critical factors. Critical success factors can still be used as assistance in the organization’s planning processes, to improve the communication (internal or external), and to assist in the development of information systems [10]. There are a number of factors beyond the control of management, and determine the success or otherwise of a project [11], these are the critical success or failure. On the other hand, the critical successes can also as resources, capabilities, and organizational attributes that are essential for performance [12]. The organizations can develop different critical success factors according to the with the structure, competitive strategy, position in industry, location, and environmental and temporal factors, integrated with the strategic objectives [13]. It’s already recognized the importance of the critical success factors since they end up influencing the achievement of the organization’s strategic objectives. According to the survey, it indicates that about 44% of IS projects are delivered out of time, above budget and without meeting the requirements [14]. In addition, about 24% are canceled before completion or are delivered but never used. These data highlight the need to overcome a multiplicity of obstacles in order to achieve results happened. The critical success factors in IS as small claims of limited number areas where “everything must run well” for the SI function to succeed [15]. The author carried out case studies nine organizations, and after the interviews the author arrived at two main conclusions: first, the critical success factors of IS different from organization to organization, however it is possible to identify a common set of four critical success factors. Secondly, each department of IS establishes its management tools, techniques and processes to achieve goals in critical areas.

344

L. W. Santoso

The set of four critical IS success factors identified in all the study organizations were: (1) the tools, techniques and processes should be in the right places in the organization so that the management of effective; (2) make sure that people are working in the correct project. With limited resources, the organization’s priorities should be carefully selected; (3) the services delivered by IS should be on time and within budget. Just as you should that the service that is delivered to you is good, otherwise the quality of the service becomes irrelevant; (4) Finally, IS priorities should be aligned with business needs and organizational goals. In turn, the concept of success in IS can be understood as the final efficiency in carrying out the task for which the information system is developed [16]. The authors set 12 case studies to small and medium-sized enterprises success in adopting IS. From the results, the authors identified two sets of factors: (1) secondary factors, which are important but not critical to success. Examples of these factors are the availability of financial resources, the quality of the software market, availability, and quality of human resources; (2) the second set of factors were classified as determinants, since clearly demonstrate why certain organizations are more successful in the adoption of IS. These factors are the IS skills (people and available knowledge in IS), and the attitudes and perspectives of the management adoption and use of IS. It indicates that approaching people, processes and technologies maximizes the chances of success in the implementation of IS, while at the same time enhancing the organizational benefits expected top management [17]. For this study we adopted interpretation of critical success factors as the essential factors to achieve success. While withdrawing value means getting insights, information and to originate knowledge relevant to the organization (e.g. a pattern of consumption; identify a need that translates into a new product; identify a threat). In this way, we intend to identify what critical success factors are most important for withdraw value from the big data. For this component of the study, analyzed scientific articles, books, content of different types, among them press releases, online articles and news.

3 Methodology and Data The main research question of this study is “what are the factors critics of success more important to withdraw value of the big data?”. Following the collection of critical success factors based on the literature review, this research proposes a qualitative exploratory study and a Delphi rankings [18], to better understand the subject. For this study we considered the Delphi methodology of rankings that uses nonparametric statistics as the most appropriate for several reasons. First of all, because it will be necessary to set out the assumptions and information that different interpretations and seek to generate consensus in a group [19]. Secondly, the Delphi methodology is an appropriate way to work with opinions and perspectives rather than objective facts [18]. Thirdly, it is suitable for studies which seek to investigate an uncertain and speculative question, where the general population is not able to questions raised by the research [20]. Finally, because it is intended to present the results in a way that their understanding is easy, making them accessible to researchers, academics, and actors in the industry. Delphi

Big Data: Identification of Critical Success Factors

345

rankings can still be advantageous for the initial phase of theoretical developments, such as the big data, since they help the researcher to identify the variables of interest and generate propositions. 3.1 First Phase The first phase focuses on the collection of information about members of the panel, and the selection and compilation of the list items [14]. The ideal panel size is ten to eighteen people and a maximum of four panels at the same time [20]. The constitution of the panel aimed at the selection of professionals who are knowledgeable, when they stated that they should be made up primarily of persons with respond to the investigator’s inquiries. The identification of these professionals was made through direct contacts with consultants and professionals in the IS sector, and academic leaders. The panel was composed by six professionals assigned to big data projects and activities, five consultants, three IS teachers, one journalist specialized in IS, a CTO of an organization specialized in an IT partner, an organization partner focused on the development and implementation of software solutions, and a director of a start up which also operates in the IT sector. Regarding the characterization of the educational qualifications, it was found that 29% had degree, 40% master’s degree, 20% doctorate and 11% postgraduate. All panelists were contacted in advance via email address, where they were explained the methodology and the objectives, to guarantee participation in this study. At this stage, it was decided to identify the main questions from the literature review. It was decided to create the list with 20 items since it seemed to be a satisfactory quantity, based on the three critical aspects of the Delphi [19]: the time must be considered for the study; the complexity and quantity of issues that the must respond and therefore panel fatigue. This is how it came to be conclusion that more items could discourage participation in this study. 3.2 Second Phase It indicates that at this stage a list with 10–20 items is maximum [18]. Therefore, it was decided to establish as objective identify the 10 critical success factors to most important date according to the panel. The list of identified and asked the panel to according to a Likert-type numerical scale [21, 22]. The values presented by the scale went from one, equivalent to nothing important, up to ten, equivalent to very important. The literature presents different views regarding the ideal size of the scales. The ideal size of the scale should reflect above all quality [23], i.e. the ability to faithfully demonstrate the attitude/opinion that is being measure. In this way it is necessary to seek a balance: scales with few classifications may lose the discriminatory capacities of the respondents; per other scales with too many ratings can go beyond the respondents’ real capacities [24]. In this study, we chose 10-class scales. The 10-point scales present the advantages when compared to the most used five and feel points: they offer greater variations; higher measurement accuracy; opportunities to detect change and powers for explanations. Larger scales also help minimize problems of benevolence, central tendency, and the “halo” effect [25].

346

L. W. Santoso

Considering the hourly availability of panel members, it was decided to use online surveys. The second phase of the study began on April 12, 2020 and ended on May 2, 2020. The questionnaires were sent to the eighteen and the response rate was 100%. 3.3 Third Phase For this phase it was decided to carry out one to three at the most because of this is the ideal number. In turn the rate of response to each round was at least 70%, since this is the minimum rate indicated by the same authors for the study to become relevant. Regarding the level of consensus, for this study the value proposed equal to 0.5 representing high agreement. However, if this value is not reached, the process should be finalized. On the other hand, the coefficient of Spearman’s correlation between rounds is intended to be the nearest possible of 1. At this stage, a specific resource was used for surveys designed to the online survey tool, which allows you to create rankings. It was then asked the panel to order the ten factors, placing first the most important and tenth or the least important. This phase started on May 07, 2020, and closed on the day May 21, 2020. The response rate was 83%. At this stage it was also SPSS tool for data processing (Table 1). Table 1. Final ordering of critical success factor (CSFs). Ranking

Critical success factor

Point

1

Aligning big data strategy to customer insights and business goals

8.2

2

Having adequately IT infrastructure

7.6

3

Having technologies and tools used to enable big data analytics processes

7.3

4

Data-driven culture in the organizations and management

6.5

5

Automated or semi-automated data analytics tool

5.8

6

Possess Business Intelligence practices

5.1

7

Ability to innovate and adapt the business model

4.7

8

The organization should allow and encourage experimentation

4.5

9

Organization should be able to implement big data scalable solutions

3.3

10

Supported decision making in automated algorithms

3.2

At this stage it is still given the hypothesis of the panel to comment or justify the ranking created. However, only the following comment was received: The option was to give relevance to the big data aspects aligned with the strategy. Without a strategy that recognizes the contribution of the big data little sense would measures. Then the tactical factors and dependent on infrastructure, for example, it makes no sense to think of scalable solutions without adequate infrastructure.

Big Data: Identification of Critical Success Factors

347

4 Analysis and Discussion After the data collection it was possible to validate that although the panel was formed by knowledgeable professionals. Having these concordance values, the study may become a reflection of the big data, which still does not seem to meet much consensus. It became clear, the importance of aligning the big approaches organizations with their strategic objectives. The big data approach should be seen as strategic rather than as a discipline within the from you. In this way it is possible to define: objectives; an eventual ROI; the impacts; the possibility of improving processes; and other important criteria for obtaining success in a big data project. The second position in both rounds was occupied by CSF is owning adequate IT infrastructure. The benefit from big data is clear, it is inevitable to improve the technological infrastructure of organization and may be reflected in changes in applications and business processes. The IT infrastructure must a low and predictable catch latency and simple queries to data; to be capable of handling large volumes of transacted volumes, many times in distributed environments; and support flexible data structures and dynamics. “Having new models and tools capable of processing large volumes of data” was the third placed in both rounds. Organizations should increase their ability to analyze new volumes of data. It is in this the need to develop new models for the vast amount of data currently generated. The fourth position was occupied by the CSF “Organizational Culture Oriented for the management of the data”. The impact of the big data, focusing on the performance of organizations oriented to the management of data. At the bottom of the top 5 was “Automated or semi-automated data analysis”. This analysis are techniques to detect patterns, identify anomalies, and extract knowledge. The news forms of computation that combine statistical analysis, processes and artificial intelligence, are able to build statistical models from large data sets. For example, Netflix uses automatic recommendation system for predicting the interests of the client comparing their movie history with a statistical model generated by from the habits of millions of other customers. The sixth place in the final round was occupied by “Possessing Business Intelligence practices (such as predictive analytics and big data antics”). More insightful organizations already recognize the importance of analytics. This due to analytics lead to a better understanding of the customer and the Marketplace. The best performing organizations had analytics efficient as a factor of differentiation. In this way the main barrier found to gain a competitive advantage from the big data was the lack of understanding how to use analytics to improve the organization—essentially capture, represent and deliver. The big size of the data sets does not automatically mean that it will be possible to extract enough information from them, only that the potential of learning is greater. It is essential to have analytical skills and in the organization to draw the necessary conclusions. The CSF “Ability to innovate and adapt the business model” is in seventh place in the ranking. Organizations are bringing together large datasets, however in many cases they do not know what to do with them. Being thus transforming this data into information and knowledge capable of competitive advantage to the organization, it will be necessary to develop new management skills and practices, and reshape the business model of organization. The new model of business, regardless of the sector of activity

348

L. W. Santoso

of the organization, should understand that capacity management and data mining will be a key factor in critical to organizational success. The CSF “Organization should allow and experimentation (e.g., rapid and small-scale tests) is in the eighth place. In the context of the big data, rapid and small tests are essential because they allow for an solutions for the entire organization. The experimentation in the big data also leads to the identification of needs, exposing variability and enabling an improvement in organizational performance. Ninth place in the ranking was occupied by CSF “Organization should be capable of implementing scalable big data solutions”. The capacities of scalability in solutions focused on data storage and management are essential today. Whether they are structured data or not structured. The adoption of big data solutions is essentially influenced by the scalability characteristics the same. Since the NoSQL (Not Only SQL) databases are an example solution capable of handling the big data, thanks to the ability to overcome the difficulties of scalability present in the relational databases. Finally, the ranking is closed by CSF “Supported Decision Making in automated algorithms”. The explosion of data in terms of volume and format has led to the need for new approaches to process and analyze large amounts of data in real time. Essentially through new algorithms or adaptations, capable of making useful insights for the organization in its decision.

5 Conclusion The present research aimed to recognize and order according to with their degree of significance and relevance, the main critical success factors for that the organization is able to withdraw value from the big data. The great motivation in choosing this topic is that the data is currently—in all sectors, in all economies, in all organizations in their business processes. The big data is still at an early stage of research. In this way it is possible to conclude that organizations and managers should not understand the technological trends. It is also necessary a strategic action for organizational adaptation to respond to new demands and to be able to withdraw value from the transformations. Data-driven decision making could promote the company’s growth. The main limitation of the present study was the degree of agreement obtained. The result of the Kendall W value was 0.32 indicating consensus moderate. This value may reflect the diversity of opinions panel, on a subject that has not yet been studied. While a round in the third phase could increase this value, it was also considering the increasing difficulty in obtaining answers and the fatigue of the panel regarding the study. However, the study offer a fundamental knowledge for practitioners as well as for future research. It would be interesting to differences between the critical success factors for withdrawing big data value and to extract value from structured data. From the obtained results it could still be valuable to try develop frameworks to ensure that the identified CSF in the study are the metrics to measure success. Finally it would be important to realize how the big data is being addressed and to see if the presented factors in the study greatly differ those that the identified reality of Indonesia organizations.

Big Data: Identification of Critical Success Factors

349

References 1. IBM: https://researcher.watson.ibm.com/researcher/view_group.php?id=4933 2. Business Wire: https://www.businesswire.com/news/home/20121022006228/en/GartnerSays-Big-Data-Creates-Big-Jobs-4.4-Million-IT-Jobs-Globally-to-Support-Big-Data-By2015 3. Santoso, L.: Cloud technology: opportunities for cybercriminals and security challenges. In: Proceedings 12th International Conference on Ubi-Media Computing, Ubi-Media, pp. 18–23, 9049539 (2019) 4. Baesens, B., Bapna, R., Marsden, J.R., Vanthienen, J., Zhao, J.L.: Transformational issues of big data and analytics in networked business. MIS Quart. (2014) 5. Jahan, T., Hasan, S.B, Nafisa, N., Chowdhury, A.A., Uddin, R., Arefin, M.S.: Big data for smart cities and smart villages: a review. In: Lecture Notes in Networks and Systems Book Series (LNNS), vol. 371 (2022) 6. Buhl, H., Heidemann, J., Moser, F., Röglinger, M.: Big data—a fashionable topic with(out) sustainable relevance for research and practice? Bus. Inf. Syst. Eng. 5(2), 65–69 (2013) 7. DeRoos, D., Deutsch, T., Eaton, C., Lapis, G., Zikopoulos, P.C.: Understanding Big Data: Analytics for Enterprise Class Hadoop and Streaming Data. McGraw-Hill, New York (2012) 8. Agrawal, D., Bernstein, P., Bertino, E., Davidson, S., Dayal, U., Franklin, M., Widom, J.: Challenges and opportunities with big data—a community white paper developed by leading researchers across the United States. http://cra.org/ccc/docs/init/bigdatawhitepaper.pdf, Mar 2012 (2012) 9. Ahmed, G.: Critical success factors and why CSFs are important for business success (2014) 10. Bullen, C.V., Rockart, J.: A Primer on Critical Success Factors. Center for Information Systems Research Working Paper, 69 (1981) 11. Belassi, W., Tukel, O.I.: A new framework for determining critical success/failure factors in projects. Int. J. Project Manage. 14(3), 141–151 (1996) 12. Lynch, R.: Corporate Strategy. Harlow. Financial Times/Prentice Hall (2003) 13. Pal, R., Torstensson, H.: Aligning critical success factors to organizational design: a study of Swedish textile and clothing firms. Bus. Process. Manag. J. 17(3), 403–436 (2011) 14. Deng, T., Keil, M., Lee, H.K.: Understanding the most critical skills for managing IT projects: a Delphi study of IT project managers. Inform. Manag. 50(7), 398–414 (2013) 15. Rockart, J.F.: The changing role of the information systems executive: a critical success factors perspective. Massachusetts Institute of Technology, Boston (1982) 16. Caldeira, M.M., Ward, J.M.: Using resource-based theory to interpret the successful adoption and use of information systems and technology in manufacturing small and medium-sized enterprises. Eur. J. Inf. Syst. 12(2), 127–141 (2003) 17. Turner, M.K.: Three Keys to IT System Success: People, Process, Technology (2014) 18. Schmidt, R.C.: Managing Delphi surveys using nonparametric statistical techniques. Decis. Sci. 28(3), 763–774 (1997) 19. Hasson, F., Keeney, S., McKenna, H.: Research guidelines for the Delphi survey technique. J. Adv. Nurs. 32(4), 1008–1015 (2000) 20. Okoli, C., Pawlowski, S.D.: The Delphi method as a research tool: an example, design considerations and applications. Inform. Manag. 42(1), 15–29 (2004) 21. Likert, R.: A technique for the measurement of attitudes. Arch. Psychol. (1932) 22. Boone, H.N., Boone, D.A.: Analyzing Likert data. J. Ext. 50(2), 1–5 (2012) 23. Munshi, J.: A method for constructing Likert scales (2014) 24. Jacoby, J., Matell, M.S.: Is there an optimal number of alternatives for Likert scale items? Study. Educ. Psychol. Measur. 31, 657–674 (1971) 25. Walker, D.H.T.: An investigation into factors that determine building construction time performance (1985)

Effect of Parameter Value of a Hybrid Algorithm for Optimization of Truss Structures Melda Yücel, Sinan Melih Nigdeli(B) , and Gebrail Bekda¸s Department of Civil Engineering, Istanbul University-Cerrahpa¸sa, 34320 Avcılar, Istanbul, Turkey {melihnig,bekdas}@iuc.edu.tr

Abstract. For optimization methodologies and also metaheuristic algorithms, there are different parameters specific-algorithm. Besides, the correct determination of the value of these parameters is an extremely important issue to reach the desired solutions optimally in a short time. In addition to this, the performance of single algorithms can be developed by combining them with different methodologies or metaheuristics. In this respect, a hybrid algorithm, which was created via a combination of flower pollination algorithm (FPA) and Jaya algorithm (JA) as metaheuristics, is beneficial to generate optimal modeling for truss structures. Also, to detect the best properties for bar members of the truss structure are targeted by using this algorithm. In this regard, the single phase of Jaya algorithm (JA) is combined with the local search phase of the flower pollination algorithm (FPA). Also, all of the metaheuristics as independent algorithms, and the hybridized one were compared and evaluated in terms of the best weight by determining various statistical measurements like mean and standard deviation values of the best weights. Keywords: Truss structures · Optimization · Metaheuristic algorithms · Flower pollination algorithm · Jaya algorithm · Hybridization · Parameter-adjusting

1 Introduction The usage of a metaheuristic algorithm is an effective way for the optimization of engineering problems and effective optimum results can be obtained. In order to increase the efficiency of the algorithm, hybrid algorithms have been also developed to increase the convergence capacity of the algorithms and prevent local optima problems. The iterative optimization process may last too long and a small improvement may be also useful in the practical application of the method. As known, metaheuristics use a metaphor, but the main concern is to solve the problem correctly. In order to make a series of similar features to develop a new algorithm, hybrid methods using the feature of existing ones may be logical. In structural engineering, hybrid methods have been used in the optimum design of steel frames [1], precast-prestressed beams [2], prestressed slab [3], tuned mass dampers [4], trusses [5–7], cantilever beams [8] and retaining walls [9, 10]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 350–358, 2024. https://doi.org/10.1007/978-3-031-50158-6_35

Effect of Parameter Value of a Hybrid Algorithm

351

In the present study, the optimization of truss structures is done via a hybrid algorithm which is a combination of Jaya Algorithm (JA) and Flower Pollination Algorithm (FPA). The parameter sensitivity of the method was also tested.

2 The Hybrid Algorithm and Parameter-Adjusting Process In the present study, a hybrid optimization algorithm is utilized for the designing of a structural truss model. This algorithm consists of the combination of flower pollination algorithm (FPA) (was developed by Xin-She Yang in 2012 [11]) and Jaya algorithm (JA) (was proposed by Rao in 2016 [12]. Furthermore, this hybridized algorithm was investigated in the previous study including optimal modeling of several structural designs [13]. In this direction, to determine the optimum solutions for the desired parameters, a specific-parameter of FPA called switch probability (sp) was evaluated. In this process, the individual performances of JA and FPA algorithms are observed and the hybrid algorithm comprised of JA-FPA is dealt with by adjusting the parameter as switch probability. So, for this case, different values of sp between 0.1 and 1.0 by increasing 0.1 with constant population, and iteration numbers as 25 and 50,000 are handled, respectively. If sp value is bigger than a random number, then the phase of JA is operated in Eq. (1).       (1) Xnew,i = Xold ,i + rand (0, 1) g b − Xold ,i  − rand (0, 1) g w − Xold ,i  The best and worst solutions are expressed via gb and gw . X new,i and X old,i are also the updated namely new, and old/existing solutions. Moreover, rand(0, 1) provides to produce a random number between (0, 1). Otherwise, the basic formulation of FPA can be seen in Eq. (2). In Eq. (3), Levy flight distribution is also described as a mathematical situation.  sp < rand (0, 1) Xold ,i + rand (0, 1)(Xj,i − Xk,i ) (2) Xnew,i = sp ≥ rand (0, 1) Xold ,i + Levy(g b − Xold ,i )    1 −1 ε−1.5 e 2ε Levy = √ (3) 2 Here, if sp value is smaller than the produced random number, the local pollination phase of FPA was applied by following Eq. (2) and added to the hybrid algorithm. Also, X j,i and X k,i reflect the two different solutions, which are selected as randomly among all of the current solutions.

3 Numerical Examples 3.1 Details of Truss Model While the optimization process is designed, a structural model as a 72-bar truss (Fig. 1) is used to evaluate the performance of the hybrid algorithm. In this respect, the major aim is also determined as the minimization of the total structural weight by optimal sizing of bar members. On the other hand, the grouping approach in terms of truss bars is not applied to the design process, too.

352

M. Yücel et al.

Fig. 1. Details of the structural model for a 72-bar truss with design properties [7].

Moreover, the design information including the parameters, and constant properties can be seen in Table 1. Here, the slenderness cases must be controlled for bar members under compression forces. For this reason, the multiple loading cases are operated on the truss model (Table 2). Also, the mentioned structural constraints are defined within Table 3 in terms of the occurred displacement and stress factors, respectively. Table 1. Some details for operating of the optimization process.

Design parameters Design constants

Property

Notation

Ranges/Values

Unit

Bar cross-section

As

0.1–3.0

in.2

Elasticity modulus

Es

107

psi

Weight per unit of volume of steel

ρs

0.1

lb/in.3

Bar number

–

72

–

Node number

–

20

–

3.2 Optimum Design for Truss Structure In this application, the objective function is determined as minimizing the total structural weight for a 72-bar truss design. While this process is realizing, the optimum parameters

Effect of Parameter Value of a Hybrid Algorithm

353

Table 2. Multiple loading conditions on nodes for a 72-bar truss model. Case

Node

Loading conditions

Unit

Px

Py

Pz

1

17

5000

5000

− 5000

2

17

0

0

− 5000

18

0

0

− 5000

19

0

0

− 5000

20

0

0

− 5000

lb/in.2

Table 3. The design constraints for a 72 bar truss structure. Structural Member

Description

Nodes All

Unit

Displacement Displacement limitation for nodes

Bars A1–72

Constraints

Stress limitation for bar members

δ < |∓0.25|

in.

Stress for compression force

Stress for tension force

σc > − 25,000

σt < + 25,000

psi

for design were tried to find, too. In this respect, to make real of this aim, both independent metaheuristics as FPA and JA together with the hybrid algorithm as JA-FPA were utilized by applying multiple populations, and iteration numbers as 25 and 500,000 together with variable values of sp parameter ranging in 0.1–1 with the increment as 0.1, respectively. In Tables 4, 5, and 6, the optimized parameters with the best weight for structure can be seen in terms of JA, FPA, and JA-FPA algorithms, respectively. Also, in Fig. 2, the changing of minimum weight values can be seen in terms of different sp values for FPA and JA-FPA. According to the results, it can be recognized that the best algorithm is JA-FPA due to the best objective function value namely minimum total weight (305.0372 lb). It can be found by differentiating with the smallest deviation namely the error value as 0.0321 lb in terms of all population members.

4 Results According to the optimization results provided by the classic version of JA and FPA, besides the hybrid algorithm as JA-FPA, which is a combined algorithm generated by benefiting of local pollination phase of FPA with classic JA, it can be noticed that JA-FPA is the most successful and effective method in terms of minimization of total structural weight at the value of 305.0372 lb by the difference as 0.0321 lb among all 25 independent candidate solutions (Table 6). Although the smallest value of standard deviation can be

354

M. Yücel et al.

Table 4. Optimization results with statistical measurements for design parameters provided by the best solution value in terms of JA. Bar

Optimum value

Bar

Optimum value

Bar

Optimum value

A1

1.8350

A25

0.7267

A49

0.1000

A2

0.1013

A26

0.1547

A50

0.1001

A3

2.9994

A27

0.1384

A51

0.1000

A4

0.1000

A28

0.7248

A52

0.1000

A5

0.1000

A29

0.1000

A53

0.1000

A6

0.6946

A30

0.6876

A54

0.1000

A7

0.1072

A31

0.1008

A55

0.2826

A8

0.7692

A32

0.1021

A56

0.6515

A9

0.7767

A33

0.1000

A57

0.6478

A10

0.1000

A34

0.1001

A58

0.4721

A11

0.6954

A35

0.1000

A59

0.8638

A12

0.1000

A36

0.1000

A60

0.1000

A13

0.1000

A37

0.5284

A61

0.6577

A14

0.1000

A38

0.3900

A62

0.1004

A15

0.1000

A39

0.8949

A63

0.1005

A16

0.1004

A40

0.2806

A64

0.6481

A17

0.1000

A41

0.1000

A65

0.1000

A18

0.1001

A42

0.7761

A66

0.8743

A19

1.6881

A43

0.1025

A67

0.1000

A20

0.1000

A44

0.7557

A68

0.1000

A21

2.2099

A45

0.7531

A69

0.1006

A22

0.1351

A46

0.1104

A70

0.1000

A23

0.7218

A47

0.7398

A71

0.8934

A24

0.1000

A48

0.1006

A72

0.1000

Minimum weight

308.3894

Mean of minimum weights

308.4156

Std. deviation of minimum weights

0.0129

sp

–

Total iteration

50000

Population

25

found with FPA for minimum weights towards all of the candidate solutions, JA-FPA is more effective to reach the best weight value. On the other side, the comment can be made that this algorithm namely the hybrid method as JA-FPA has an extremely stabile

Effect of Parameter Value of a Hybrid Algorithm

355

Table 5. Optimization results with statistical measurements for design parameters provided by the best solution value in terms of FPA. Bar

Optimum value

Bar

Optimum value

Bar

Optimum value

A1

1.8917

A25

0.7961

A49

0.1011

A2

0.1019

A26

0.1097

A50

0.1000

A3

2.2641

A27

0.1068

A51

0.1006

A4

0.1106

A28

0.7683

A52

0.1001

A5

0.1004

A29

0.1046

A53

0.1008

A6

0.7083

A30

0.7255

A54

0.1002

A7

0.1103

A31

0.1001

A55

0.2576

A8

0.8000

A32

0.1001

A56

0.4132

A9

0.8221

A33

0.1000

A57

0.5457

A10

0.1006

A34

0.1004

A58

0.3957

A11

0.6964

A35

0.1001

A59

0.8982

A12

0.1048

A36

0.1002

A60

0.1000

A13

0.1001

A37

0.5213

A61

0.7076

A14

0.1003

A38

0.2660

A62

0.1000

A15

0.1000

A39

0.7169

A63

0.1014

A16

0.1001

A40

0.2970

A64

0.6309

A17

0.1002

A41

0.1019

A65

0.1009

A18

0.1000

A42

0.8185

A66

0.8573

A19

1.6229

A43

0.1073

A67

0.1004

A20

0.2925

A44

0.7586

A68

0.1000

A21

2.0611

A45

0.7768

A69

0.1001

A22

0.2576

A46

0.1014

A70

0.1037

A23

0.7223

A47

0.8002

A71

0.8911

A24

0.1001

A48

0.1001

A72

0.1001

Minimum weight

305.3980

Mean of minimum weights

305.4480

Std. deviation of minimum weights

0.0283

sp

0.8

Total iteration

50000

Population

25

behavior in the way of providing the best amount of the minimized weight towards each sp value in comparison to FPA (Fig. 2).

356

M. Yücel et al.

Table 6. Optimization results with statistical measurements for design parameters provided by the best solution value in terms of JA-FPA. Bar

Optimum value

Bar

Optimum value

Bar

Optimum value

A1

1.7761

A25

0.7543

A49

0.1003

A2

0.1045

A26

0.1233

A50

0.1009

A3

2.2557

A27

0.1296

A51

0.1001

A4

0.1000

A28

0.7925

A52

0.1000

A5

0.1005

A29

0.1027

A53

0.1000

A6

0.7309

A30

0.7351

A54

0.1002

A7

0.1013

A31

0.1003

A55

0.3094

A8

0.8001

A32

0.1000

A56

0.4120

A9

0.7964

A33

0.1000

A57

0.5644

A10

0.1000

A34

0.1000

A58

0.4112

A11

0.7411

A35

0.1000

A59

0.8855

A12

0.1057

A36

0.1000

A60

0.1003

A13

0.1000

A37

0.4034

A61

0.6993

A14

0.1000

A38

0.2826

A62

0.1004

A15

0.1001

A39

0.7022

A63

0.1000

A16

0.1000

A40

0.3117

A64

0.6237

A17

0.1000

A41

0.1000

A65

0.1050

A18

0.1000

A42

0.7858

A66

0.9056

A19

1.6487

A43

0.1000

A67

0.1011

A20

0.2611

A44

0.7404

A68

0.1001

A21

2.1146

A45

0.7519

A69

0.1000

A22

0.2476

A46

0.1005

A70

0.1010

A23

0.7581

A47

0.7834

A71

0.8963

A24

0.1039

A48

0.1005

A72

0.1000

Minimum weight

305.0372

Mean of minimum weights

305.1410

Std. deviation of minimum weights

0.0321

sp

0.8

Total iteration

50000

Population

25

Effect of Parameter Value of a Hybrid Algorithm

357

Fig. 2. Minimum structural weight according to variable sp values.

5 Conclusion As a summarization, JA-FPA hybrid algorithm is the most effective method to minimize the total structural weight in terms of 72-bar truss model among the other algorithms as classical versions of JA FPA. Also, this algorithm can make a convergence to the best amounts of weights with extremely low standard deviations of objective functions for each candidate solution and sp values. Besides these results, it is clear that the best sp value is determined as 0.8 in terms of reaching the minimized value of weight by both FPA and hybrid algorithm as JA-FPA.

References 1. Akin, A., Aydogdu, I.: Optimum design of steel space frames by hybrid teaching-learning based optimization and harmony search algorithms. World Acad. Sci. Eng. Technol. Civ. Environ. Eng. 2(7), 739–745 (2015) 2. Yepes, V., Martí, J.V., García-Segura, T.: Cost and CO2 emission optimization of precast– prestressed concrete U-beam road bridges by a hybrid glowworm swarm algorithm. Autom. Constr. 49, 123–134 (2015) 3. Talaei, A.S., Nasrollahi, A., Ghayekhloo, M.: An automated approach for optimal design of prestressed concrete slabs using PSOHS. KSCE J. Civ. Eng. 21(3), 782–791 (2017). https:// doi.org/10.1007/s12205-016-1126-9 4. Nigdeli, S.M., Bekda¸s, G., Yang, X.S.: Optimum tuning of mass dampers by using a hybrid method using harmony search and flower pollination algorithm. In: International Conference on Harmony Search Algorithm, pp. 222–231. Springer, Singapore (2017) 5. Panagant, N., Bureerat, S.: Truss topology, shape and sizing optimization by fully stressed design based on hybrid grey wolf optimization and adaptive differential evolution. Eng. Optim. 50(10), 1645–1661 (2018) 6. Omidinasab, F., Goodarzimehr, V.: A hybrid particle swarm optimization and genetic algorithm for truss structures with discrete variables. J. Appl. Comput. Mech. 6(3), 593–604 (2020) 7. Bekda¸s, G., Yucel, M., Nigdeli, S.M.: Evaluation of metaheuristic-based methods for optimization of truss structures via various algorithms and Lèvy flight modification. Buildings 11(2), 49 (2021)

358

M. Yücel et al.

8. Toklu, Y.C., Bekda¸s, G., Kayabekir, A.E., Nigdeli, S.M., Yücel, M.: Total potential optimization using metaheuristics: analysis of cantilever beam via plane-stress members. In: International Conference on Harmony Search Algorithm, pp. 127–138. Springer, Singapore (2020) 9. Yücel, M., Kayabekir, A.E., Bekda¸s, G., Nigdeli, S.M., Kim, S., Geem, Z.W.: Adaptive-hybrid harmony search algorithm for multi-constrained optimum eco-design of reinforced concrete retaining walls. Sustainability 13(4), 1639 (2021) 10. Sharma, S., Saha, A.K., Lohar, G.: Optimization of weight and cost of cantilever retaining wall by a hybrid metaheuristic algorithm. Eng. Comput., 1–27 (2021) 11. Yang, X.S.: Flower pollination algorithm for global optimization. In: International Conference on Unconventional Computing and Natural Computation, pp. 240–249. Springer, Berlin, Heidelberg (2012) 12. Rao, R.: Jaya: a simple and new optimization algorithm for solving constrained and unconstrained optimization problems. Int. J. Ind. Eng. Comput. 7(1), 19–34 (2016) 13. Yücel, M., Bekda¸s, G., Nigdeli, S.M.: Optimization of truss structures with sizing of bars by using hybrid algorithms. In: International Conference on Intelligent Computing and Optimization, pp. 592–601. Springer, Cham (2021)

Brain MRI Classification for Alzheimer’s Disease Based on Convolutional Neural Network Md. Saiful1 , Arpita Saha1 , Faria Tabassum Mim1 , Nafisa Tasnim1 , Ahmed Wasif Reza1(B) , and Mohammad Shamsul Arefin2,3(B) 1 Department of Computer Science and Engineering, East West University, Dhaka 1212,

Bangladesh [email protected] 2 Department of Computer Science and Engineering, Daffodil International University, Dhaka 1216, Bangladesh [email protected] 3 Department of Computer Science and Engineering, Chittagong University of Engineering and Technology, Chattogram 4349, Bangladesh

Abstract. Alzheimer’s disease is a severe disorder of the brain that gradually increases and affects the function of the brain. It mainly affects middle-aged people or old aged person. Many researchers tried to train their model to classify or detect Alzheimer’s disease from MRI images automatically. In this paper, we also tried to classify four classes (Mild Demented, Moderate Demented, NonDemented, Very Mild Demented) of Alzheimer’s diseases using ResNet (Residual neural network) on 6400 MRI images. In the paper, ResNet50v2 and ResNet101v2 used. By comparing their performance, ResNet101v2 gave a better result. The model’s precision is 74%, 27%, 75%, and 54%, recall percentage is 28%, 25%, 65%, and 77%, and f1 scores are 40%, 26%, 70%, and 63% for mild demented, moderate demented, non-demented, and very mild demented, respectively. By applying ResNet101v2, the percentage of accuracy is 98.35%. Keywords: Brain MRI · Alzheimer’s disease · Deep learning · ResNet · CNN

1 Introduction Alzheimer’s disease (AD) is the type of dementia that is most prevalent. Minor memory loss is where it starts, and it might eventually proceed to communication and environmental awareness loss. One in every 100 persons has Alzheimer’s disease, an incurable brain disorder [1]. Alzheimer’s disease typically impacts individuals who are close to 65 years old, although it can also affect young people in rare cases. Approximately 10.7% of individuals aged 65 and above have this condition. The likelihood of having Alzheimer’s increases with age, with 5.0% of people aged 65–74, 13.1% of people aged 75–84, and 33.2% of people aged 85 or older being affected by the disease [2]. All forms of Alzheimer’s disease go through four stages: preclinical, early, middle, and late [3]. In the past, preclinical Alzheimer’s disease was described as a condition where individuals had brain abnormalities associated with the disease despite having normal cognitive function during their lifetime. This was typically identified upon © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 359–367, 2024. https://doi.org/10.1007/978-3-031-50158-6_36

360

Md. Saiful et al.

examination of the brain after the person’s death [4]. Years before a person exhibits any symptoms of the disease, preclinical stage alterations in the brain start. The most prevalent symptoms in the early stages include forgetfulness, such as forgetting a recent event or activity. Alzheimer’s disease can affect the brain in its middle stage, making it difficult to speak and do daily tasks. A person can also lose their memory for two to four years, repeating the same error in a different circumstance, and so on. In the last stage, Alzheimer’s patient finally loses the capacity to speak and care for themselves. A person loses their capacity to identify, eat, go to the bathroom, and perform other essential functions. They become physically unable to sit, walk, and do other activities. Histological findings explain that AD is characterized by hyperphosphorylated tau protein deposition in internal neurofibrillary tangles and excessive amyloid protein deposition in exterior plaques, which can affect neuronal mortality. The hemispherical structural uniformity of the human brain is highly developed and declines with age. Pathological diseases, such as AD, have a more significant negative impact on this symmetry. Regional grey matter loss frequently characterizes the anatomical hemispherical asymmetrical progression in AD, with the left hemisphere regions being more severely and firstly affected by the degeneration process. To create biomarkers for Alzheimer’s disease (AD), various research has looked at the extent of the symmetric brain loss in Magnetic Resonance (MRI) Images, particularly of the temporal areas. Shi performed a meta-analysis of MRI studies to assess shrinkage and inequity patterns in MCI and AD. They discovered that MCI had lower bilateral hippocampal shrinkage and atrophy than AD did and that all groups shared a consistent pattern of left-less-than-right asymmetry, albeit varying degrees. Knowledge of anatomy, anatomical landmarks and sectoral charts, and Machine Learning and Artificial Intelligence structures that execute learning and memory analysis or study the entire brain have been designed to assist decision-making [5]. Advanced machine learning techniques have made it possible to detect the presence of diseases with greater reliability through the use of Positron Emission Tomography (PET), Computed Tomography (CT), and Magnetic Resonance Imaging (MRI). These imaging technologies rely on complex computational algorithms to theoretically identify the existence of diseases. The anatomical atrophic alterations in the brain are visible on brain MRIs. The computerized brain MRI analysis benefited greatly from machine learning experiments employing neuroimaging data. One of the most helpful machine learning techniques for diagnosing Alzheimer’s disease is deep learning (DL). DL models can be used to extract discriminative features from raw data automatically. The most sophisticated DL architectures are created for image segmentation, regression, and classification using real-world images. To learn all characteristics encoded in images, these models need a lot of data, such as the report of an MR scan of the brain. The advantage of using DL models is that they do not require manually generated AD features because learned features are immediately taken from input photos [6]. Our main goal behind this research is to classify AD from brain MRI Images and implement deep learning for analyzing data. Moreover, we want to analyze the model’s efficiency for AD classification. We are trying to show the comparison between ResNet50v2 and ResNet101v2.

Brain MRI Classification for Alzheimer’s Disease

361

2 Literature Review Alinsaif tried classifying AD using 3D-ST and CNN. Here, they proposed a system that consists of two pipelines. In the first step, they worked with MRI samples. In the second step, they used CNN models. Applying the 3D-ST, they obtained 62–68% accuracy [7]. Acharya, Mehta, and Kumar worked with CNN, VGG16, ResNet, and AlexNet. They mainly tried to classify AD into four categories. Transfer learning helps to reduce training time. By using CNN, VGG16, and ResNet50, they got better results than other models. The accuracy of the model is 95.70% [3]. Tuan, Bao, Kim, Manuel, and Tavares used 3D brain MRI to diagnose AD. They got 0.96 accuracies for segmentation as well as 0.88 for classification. In order to segment, or identify distinct regions or structures, GMM is employed. To classify Alzheimer’s disease, a Support Vector Machine is utilized [8]. Ebrahimi and Luo trained MRI slices in three ways: in single-view mode using a 2D CNN, in multi-view and single-view modes using an LSTM model, and in 3D CNN mode using an MRI volume. Here, different MRI-based AD detection methods employing a 2D CNN methodology were employed to transfer learning between SqueezeNet and ResNet-18 datasets. CNN significantly enhanced the findings with transfer learning, reaching 96.88% accuracy [9]. Murugan applied deep learning techniques, and Alzheimer’s disease is categorized in MRI scans using the CNN model. They used six layers: input layer, convolution layers, pooling layers, DEMNET block, dropout layers, and Dense layers. The accuracy of DEMNET was 95.23%, and AUC was 97% [10]. Odusami, Maskeli¯unas, and Damaševiˇcius used ResNet18 and DenseNet201 models. A technique known as a gradient class activation map was utilized to identify the specific area of the MRI image. After applying ResNet18 and SVM, they got 78.72%. On the other hand, they got 98.21% accuracy for using ResNet18 and DenseNet201 with weights. Comparing the two models, they got better results in ResNet18 and DenseNet201 [11]. The works in [12–16] also focused on image analysis for performing different important tasks. After analyzing some papers, different researchers used different models like CNN, VGG, ResNet, AlexNet, DEMNET, DenseNet, LSTM, and SqueezeNET. By observing these models’ performance, we proposed ResNet50v2 and ResNet101v2 models to classify AD from MRI scans.

3 Methodology The workflow of our suggested model is shown in Fig. 1. Images are first loaded and then go through some important pre-processing steps. After that, the dataset is divided for training and testing. 3.1 Dataset For this research, we fetched our dataset from Kaggle [17] because the range of the dataset is fair enough. We did not follow any process to handle this dataset because it is

362

Md. Saiful et al.

Fig. 1. A workflow diagram for the proposed model

already balanced. The entire dataset consisted of 6400, The data was divided into two sets for training and testing purposes. There were a total of 717, 52, 2560, and 1792 images included in the training set for each category of Mild, Moderate, Non, and Very Mild Demented. Additionally, there were 179, 12, 640, and 448 test images for each of the respective categories. However, some categories, like Moderately Demented, were represented less frequently than others. To account for this, data augmentation was used during the training process to increase class balance. Some images are given in Fig. 2 from our dataset.

Fig. 2. Sample images from the dataset

3.2 Preprocessing In order to achieve optimal accuracy when training a model using the provided dataset, it is necessary to preprocess the images as they are not uniform in size. This involves resizing all images to a standard size of 10 × 10 × 1 pixels. Additionally, data augmentation is utilized during the training process to improve model performance and accuracy in predicting results. Failure to preprocess the dataset in this way would likely result in lower accuracy when training the model.

Brain MRI Classification for Alzheimer’s Disease

363

3.3 Proposed Model The focus of this study was to classify Alzheimer’s disease using MRI images as data. A Convolutional Neural Network (CNN) was implemented for feature extraction and classification, with the MRI slices being used as input. The CNN model served as the classifier, grouping each image into one of four categories: Mild, Moderate, Non, and Very Mild Demented. A ResNet-101v2 and ResNet-50v2 model was used to classify CNN blocks. The summary of ResNet-50v2: Layer(type)

Output shape

Param #

input_2(InputLayer) rescaling(Rescaling) resnet50v2(Functional) global_average_pooling2d(globalAveragePooling2D) dropout(Dropout) dense(Dense)

[(None, 176, 208, 3)] (None, 176, 208, 3) (None, 6, 7, 2048) (None, 2048) (None, 2048) (None, 4)

0 0 23564800 0 0 8196

In total params: 23,572,966 In trainable params: 18,068,996 In non-trainable params: 5,504,000

The summary of ResNet-101v2: Layer(type)

Output shape

Params

input_4(InputLayer) rescaling(Rescaling) resnet101v2(Functional) global_average_pooling2d_1(GlobalAveragePooling2D) dropout_1(Dropout) dense_1(Dense)

[(None, 176, 208, 3)] (None, 176, 208, 3) (None, 6, 7, 2048) (None, 2048) (None, 2048) (None, 4)

0 0 42626560 0 0 8196

In total params: 42,634,756 In trainable params: 8196 In non-trainable params: 42,626,560

3.4 Architecture ResNet-101 (Residual Neural Networks) are used in the architecture for classification applications. In ResNet design, The (1 × 1) layers perform dimensionality reduction and restoration before and after the (3 × 3) layer. ResNet-101 architecture includes information on the building elements and how many blocks are stacked. In Fig. 3, the convolutions of conv3 1, conv4 1, and conv5, and the image obtained are down-sampled. The bottleneck blocks are unique for each stage.

364

Md. Saiful et al.

Fig. 3. Architecture of the proposed model

4 Experimental Results In our proposed model, we used ResNet50v2 and ResNet101v2 in CNN architecture. For ResNet50v2 and ResNet101v2, we used two types of epochs to evaluate our models: epoch 50 and epoch 70, respectively. The accuracy was good for ResNet101v2 (epoch 70). Table 1 shows the accuracy percentage after applying ResNet101v2 (50 epochs vs. 70 epochs). In the same way, we include Fig. 4 to visualize the accuracy percentage in graphs. Table 1. Accuracy of ResNet101v2 Class

ResNet101v2 Epoch 50 (accuracy) (%)

Epoch 70 (accuracy) (%)

Mild-demented

61.64

100

Moderate-demented

33.98

100

Non-demented

66.81

99.24

Very-mild-demented

36.13

94.14

Table 2 shows the accuracy percentage after applying ResNet50v2 (50 epochs vs. 70 epochs). In the same way, we include Fig. 5 to visualize the accuracy percentage in graphs. 4.1 Performance Evaluation We have evaluated our models using different metrics such as precision, recall, and f1-score, which are represented by Eqs. (1), (2), and (3) respectively. To determine the overall accuracy of our model, we used Eq. (4). We have presented the percentage values of precision, recall, F1-score, support, and accuracy for the ResNet101v2 model in Table 3. precision of the model =

TP TP + FP

(1)

Brain MRI Classification for Alzheimer’s Disease

365

Fig. 4. Model accuracy and loss for ResNet101v2

Table 2. Accuracy of ResNet50v2 Class

ResNet50v2 Epoch 50 (accuracy) (%)

Epoch 70 (accuracy) (%)

Mild-demented

51.28

100

Moderate-demented

25.63

99.85

Non-demented

54.21

99.14

Very-mild-demented

29.69

92.5

Fig. 5. Model accuracy and loss for ResNet50v2

recall of the model = F1 score of the model =

TP TP + FN

(2)

2(precision ∗ recall) (precision + recall)

(3)

correct prediction all predictions

(4)

Accuracy of the model =

366

Md. Saiful et al. Table 3. Performance evaluation of the ResNet101v2 model

Alzheimer

Precision (%)

Recall (%)

F1-score (%)

Support

Accuracy (%) 98.35

Mild-demented

74

28

40

179

Moderate-demented

27

25

26

12

Non-demented

75

65

70

640

VeryMD

54

77

63

448

To train this dataset, we used ResNet50v2 and ResNet101v2. ResNet models are used in the research because it is one of the most powerful models which can detect AD from MRI scans. From ResNet101v2, we got better results comparatively. Figures 4 and 5 show the model’s accuracy and loss for 50 and 70 epochs, respectively. In ResNet101v2, 377 layers are used in this model. Here the accuracy of 50 epochs is 49.64%. Similarly, the accuracy of 70 epochs is 98.35%. In Fig. 4, there are also decreases for 70 epochs exponentially. In the ResNet50v2 model, 190 layers are used. Here the accuracy of 50 epochs is 40.20%. Similarly, the accuracy of 70 epochs is 97.87%. In Fig. 5, loss accuracy is decreased for 50 epochs. From some previous discussions, our model accuracy was pretty good from the other models. However, some researchers applied to VGG-16 and ResNet50v2, for which they got 95.70%, whereas after applying Resnet101v2, we got 98.35% accuracy. For this reason, this model can be used to classify AD.

5 Conclusion In this paper, we used the ResNet architecture to identify four classes of Alzheimer’s disease. We use 377 layers in the ResNet101v2 model and 190 in the ResNet50v2 model. The accuracy of ResNet50v2 is 97.87%, and the accuracy of ResNet101v2 is 98.35% (proposed model). By comparing RestNet50v2 and ResNet101v2, ResNet101v2 gets better accuracy. In the future, there will be a greater emphasis on categorizing and identifying areas in Positron Emission Tomography (PET) and Functional Magnetic Resonance Imaging (FMRI) scans related to Alzheimer’s Disease, along with region detection.

References 1. Subramoniam, M., Aparna, T.R., Anurenjan, P.R., Sreeni, K.G.: Deep learning based prediction of Alzheimer’s disease from magnetic resonance images [Online] (2021). Available: http://arxiv.org/abs/2101.04961 2. Wong, S., et al.: Editorial Board Editor-in-Chief [Online] (2023). Available: www.elsevier. com/locate/compmedimag 3. Acharya, H., Mehta, R., Kumar Singh, D.: Alzheimer disease classification using transfer learning. In: Proceedings—5th International Conference on Computing Methodologies and Communication, ICCMC 2021, April 2021, pp. 1503–1508. http://doi.org/10.1109/ICCMC5 1019.2021.9418294

Brain MRI Classification for Alzheimer’s Disease

367

4. Hubbard, M., Fentont, G.W., Anderson, J.M.: A quantitative histological study of early clinical and preclinical Alzheimer’s disease (1990) 5. Poloni, K.M., Duarte de Oliveira, I.A., Tam, R., Ferrari, R.J.: Brain MR image classification for Alzheimer’s disease diagnosis using structural hippocampal asymmetrical attributes from directional 3-D log-Gabor filter responses. Neurocomputing 419, 126–135 (2021). https:// doi.org/10.1016/j.neucom.2020.07.102 6. Sh Aaraji, Z., Abbas, H.H.: Automatic classification of Alzheimer’s disease using brain MRI data and deep convolutional neural networks (2022) 7. Alinsaif, S., et al.: 3D shearlet-based descriptors combined with deep features for the classification of Alzheimer’s disease based on MRI data ARTICLEINFO [Online] (2021). Available: http://adni.loni.usc.edu/wp-content/ 8. Tuan, T.A., The Bao, P., Kim, J.Y., Manuel, J., Tavares, R.S.: Alzheimer’s diagnosis using deep learning in segmenting and classifying 3D brain MR images (2022) 9. Ebrahimi, A., Luo, S., Alzheimer’s Disease Neuroimaging Initiative: Convolutional neural networks for Alzheimer’s disease detection on MRI images. J. Med. Imaging 8(02) (2021). http://doi.org/10.1117/1.jmi.8.2.024503 10. Murugan, S., et al.: DEMNET: a deep learning model for early diagnosis of Alzheimer diseases and dementia from MR images. IEEE Access 9, 90319–90329 (2021). https://doi. org/10.1109/ACCESS.2021.3090474 11. Odusami, M., Maskeli¯unas, R., Damaševiˇcius, R.: An intelligent system for early recognition of Alzheimer’s disease using neuroimaging. Sensors 22(3) (2022). http://doi.org/10.3390/s22 030740 12. Saha, R., Debi, T., Arefin, M.S.: Developing a framework for vehicle detection, tracking and classification in traffic video surveillance. In: Vasant, P., Zelinka, I., Weber, G.W. (eds.) Intelligent Computing and Optimization. ICO 2020. Advances in Intelligent Systems and Computing, vol. 1324. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-681548_31 13. Fatema, K., Ahmed, M.R., Arefin, M.S.: Developing a system for automatic detection of books. In: Chen, J.IZ., Tavares, J.M.R.S., Iliyasu, A.M., Du, K.L. (eds.) Second International Conference on Image Processing and Capsule Networks. ICIPCN 2021. Lecture Notes in Networks and Systems, vol. 300. Springer, Cham (2022). https://doi.org/10.1007/978-3-03084760-9_27 14. Rahman, M., Laskar, M., Asif, S., Imam, O.T., Reza, A.W., Arefin, M.S.: Flower recognition using VGG16. In: Chen, J.IZ., Tavares, J.M.R.S., Shi, F. (eds.) Third International Conference on Image Processing and Capsule Networks. ICIPCN 2022. Lecture Notes in Networks and Systems, vol. 514. Springer, Cham (2022). http://doi.org/10.1007/978-3-031-12413-6_59 15. Yeasmin, S., Afrin, N., Saif, K., Imam, O.T., Reza, A.W., Arefin, M.S.: Image classification for identifying social gathering types. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). http://doi.org/10.1007/ 978-3-031-19958-5_10 16. Ahmed, F., et al.: Developing a classification CNN model to classify different types of fish. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). http://doi.org/10.1007/978-3-031-19958-5_50 17. “Alzheimer’s Dataset (4 Class of Images).” | Kaggle, https://www.kaggle.com/datasets/touris t55/alzheimers-dataset-4-class-of-images

Drivers and Barriers for Going Paperless in Tertiary Educational Institute Rafid Mahmud Haque1 , Lamyea Tasneem Maha1 , Oshin Nusrat Rahman1 , Noor Fabi Shah Safa1 , Rashedul Amin Tuhin1 , Ahmed Wasif Reza1(B) , and Mohammad Shamsul Arein2,3(B) 1 Department of Computer Science and Engineering, East West University, Dhaka, Bangladesh

{2019-1-60-085,2019-1-60-055,2019-1-60-014, 2019-1-60-060}@std.ewubd.edu, {mcctuhin,wasif}@ewubd.edu 2 Department of CSE, Daffodil International University, Dhaka, Bangladesh [email protected] 3 Department of CSE, Chittagong University of Engineering and Technology, Chattogram, Bangladesh

Abstract. Tertiary educational institutes especially universities have become one of the major consumers of paper. Since paper uses natural resources and energy, universities are being encouraged to turn processes paperless. However, there are certain barriers along with drivers which lead to turning processes paperless. Although some work has been done on similar work, most of these works are not from Bangladesh’s perspective and do not show as-is or to-be models of processes. This paper presents the current scenario of paperless initiatives, the impact of going paperless, and the driving forces and barriers to making processes paperless. Due to limitations in time and resources, only personnel working at East West University were interviewed for this qualitative study. To analyze the data from the interviews, the ‘Gap Analysis’ method. The interview data was then used to build as-is and to-be models for two processes at East West University which are not yet paperless. The drivers and barriers to these processes turning paperless were also explored. Going paperless not only reduces paper usage, but also reduces manpower, physical storage space, and time needed to maintain the processes. However, barriers specific to different processes still remain which include but are not limited to fear of data center failure and transparency maintenance. Keywords: Paperless · Drivers · Barriers · Process model

1 Introduction In a world full of modern technology, going paperless is a sustainable form of practice. Going paperless – as the name describes is a process to reduce the usage of paper and shift to digital documentation. Going paperless is time efficient, easy to store, and lastly environment friendly. As papers are made from fiber that comes from trees, yearly, millions of trees are cut down © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 368–378, 2024. https://doi.org/10.1007/978-3-031-50158-6_37

Drivers and Barriers for Going Paperless in Tertiary

369

for making paper. Going paperless can reduce the number of trees cut and hence help to be more environmentally friendly. Organizations use a lot of paper to store information. As a result, a lot of trees are cut down to meet the required amount of paper used by the current population resulting in environmental damage [1]. Moreover, using papers lead to slow transfer or loss of information and low data security [1]. Also, manufacturing papers use water, trees, and energy which is environmentally expensive [2]. In fact, if resource preservation, pollution prevention, and cost reduction are considered, waste minimization is more beneficial compared to recycling [2]. Hence, organizations must focus on reducing paper consumption rather than on ways to deal with the used paper. To solve these issues, the concept of going paperless emerged in 1975 for offices that were to use computers instead of paper to store information [3]. Now, when educational institutes, especially universities, have become one of the largest consumers of paper usage, they are being encouraged to reduce the amount of paper used. Examples of institutes that have realized the benefits of turning processes paperless include Yale University, the University of California, and Xavier University [3]. The more papers used in tertiary educational institutes, the more resources, like trees and energy, will be used. Since the first emergence of the idea of turning organizations paperless back in 1975, many universities have been trying to turn organizational processes paperless. There have not been many empirical studies to know why universities in Bangladesh are not implementing paperless environments at the service level and if some universities did implement such environments what were the drivers and barriers. This study will focus on the following objectives 1. To identify which processes can be paperless. 2. To identify the process’s as-is model and to-be model. 3. Find out the drivers and barriers of identified processes. This study will focus on finding the drivers and barriers to making processes paperless in tertiary educational institutes in Bangladesh and drawing the as-is and to-be models of those processes. This study will only be focusing on one departmental (departmental archiving of the Computer Science and Engineering department) and one administrative (graduation application) process at East West University.

2 Related Work Similar work has been done related to this field in the past but most of those works are not empirical and not from Bangladesh’s perspective. In one of the studies [4], the sectors which are trying to implement paperless applications in Indonesia were identified and compared with those that are moving most towards paperless. The authors collected data from the education, government, and industry sectors and found that the education sector uses the most paperless applications. In another study [5], the authors described the practices and methods that minimize paper usage. In order to reduce the usage of paper and enhance the system workflow, these existing problems need to be solved. The authors mentioned the government initiatives taken at the institutional level, ECM guidelines, etc. for making paperless administration.

370

R. M. Haque et al.

In another study [6], the negative impacts of paper-based systems on the environment which include, increasing deforestation and hindering sustainability, are mentioned. The paper-based system led to harmful effects on the environment. After the findings, the authors came to the conclusion that 325 people from the paperless campus could protect 195 trees. To conclude, the paper recommends that universities should embrace paperless campuses for ensuring a sustainable low carbon society. A study [3] was carried out to find paperless models in Uzbekistan universities. Paper usage there led to the loss of information, duplication of papers, and corruption. The paper suggests services like timetable management for students, teachers, and transportation, online appointment, and usage of E-student cards. In another study [1], the barriers to the implementation of paperless processes in Nigerian universities were described. Paper-based processes were slow and unreliable. The suggested solutions include getting political actors’ financial support, filling the infrastructural gap in the ICT area, training staff members, making electronics and the internet affordable, and replacing old computers with powerful ones. Another study [7] was carried out to find out the writing and reading preferences of students in a paperless classroom. The authors found that students prefer to read short texts or articles on computers but, prefer to read longer texts or stories in printed papers. The authors also found that students prefer to use the computer for long texts or assignments rather than writing on paper. In one study [8], authors attempted to seek the possibility of a paperless Higher Education system in Nigeria which sought to unravel the concept, challenges, and prospects of the paperless school system as a means to solving the current challenges of record keeping in their school system. The study established that the paperless system is a 21stcentury trend and should be embraced by a higher system of knowledge and identified areas to begin the paperless system which can be achieved with committed efforts by management, decision makers/implementation personnel, and the general school system. In another study [9], the theory and practice of paperless classrooms are analyzed critically, and an observational analysis of the undergraduate students of Dhofar University and their engagement with digital devices in the classroom has been done. The study finds that paperless classrooms are more dynamic, engaging, and productive will enable learners not only to develop autonomous learning, data collection, and analysis, collaboration, and teamwork in the classroom but also delimit the geographical and time restrictions which can be done by Paperless classrooms equip learners with the technological skills. The other study [10] aims to investigate how capable, and available the students at senior high schools are at implementing paperless classrooms with substitutes such as the digital mode in learning activities. For the amount of 108 students majoring in Computer Network Engineering in the Bekasi area from different grades are chosen as a sample and found that there are significant influences from Environmental Awareness, digital literacy, and habit affected the student’s readiness to implement the paperless Concept of 41.3%. Another study [11] was carried out to find the student’s perception regarding paperless English classrooms at Japan IT University. The authors find out that, the major 4 concerning factors for a paperless classroom are - skill, tools, vocabulary, and notes.

Drivers and Barriers for Going Paperless in Tertiary

371

The authors also described that the transition from paper to a paperless classroom won’t be smooth. The instructors of the university have to show students, how to use digital materials for academic purposes.

3 Methodology 3.1 Data Collection Method The method of collecting data is one of the most necessary parts of the research process. In order to investigate qualitative research, some examples of data collection methods are interviews, focus groups, and observation [12]. The interview was chosen as the data collection method so that data regarding the processes can be collected from the chosen experienced people. For the research purpose, the authors have chosen to take interviews in the semi-structured form which is preferable since it offers a great deal of flexibility to the interviewer. 3.2 Participants and Sampling With the given time, the authors were able to take 5 interviews. The 5 participants of the interview were from East West University and were interviewed to know about 2 processes; graduation application and departmental archiving. The participants for the interview were selected based on two factors; the position they are working in and the number of years they worked at the university. To know about the graduation application process, the authors interviewed one person from the Controller of Examination, who is also involved with policy-making for this process, and another personnel who is in charge of overviewing this process. To know about departmental archiving, the authors interviewed 2 senior faculties and in charge personnel who have been in the CSE department for a long time and thus have good organizational memory. 3.3 Data Analysis For analyzing the gathered data from interviews, the authors chose the “Gap Analysis” method. This method is a combination of narrative and grounded theory analysis. Here, 4 major steps which lead to the answers to the research questions are [13]: • Identify the present scenario: From the interviews, authors identified 2 processes that can be paperless. Authors have drawn their as-is process using BPMN. • Determine the ideal scenario: Authors have drawn the to-be model of the 2 processes using the interviewee’s suggestions. • Highlight the gaps that exist: Authors identified the major barriers from data. • Plans to fill the gaps: Authors extracted the ideas to implement the to-be model.

372

R. M. Haque et al.

3.4 Research Ethics The major ethical issues that need to be considered include informed consent, protection of the confidentiality of the participants, and respect for the participants’ privacy. A consent form was provided to each participant of the interview which was to be read by them before the researcher could take any interview. All the data that will be used, will be kept confidential and anonymous. No data will be used for the malicious purpose which may result in any kind of distresses to the participants.

4 Results As mentioned, qualitative data for this study was collected through interviews. The authors interviewed each of the 2 faculty members twice. From the first interview, they gathered information on processes that could be paperless. From the different processes mentioned, they chose to focus on departmental archiving and the graduation application process. In a second interview with the faculties, the authors specifically concentrated on the chosen two processes. The authors also interviewed 3 personnel related to these two processes. The collected data were analyzed to find an estimate of paper usage and drivers and barriers to those processes. From interviews authors took from the CSE department and Controller of Examination office, authors found that 2 major processes can go paperless which are the graduation application process and departmental archiving. The current models of these 2 processes are very complex and are time and paper-consuming. The as-is model of the processes is given below

Fig. 1. As-is model of the graduation application process

In Fig. 1, we see that the graduation application process starts with the student making a payment. After this, the student must come to the university and submit different documents like grade reports, application forms, and course checklists. These documents

Drivers and Barriers for Going Paperless in Tertiary

373

are then sent to the Exam Office, Registrar’s Office, Library, Accounts, and finally to the respective department of the student for clearance. After the final clearance, the documents are then sent to the Exam office. If all the clearances are successful, the grade report is stored in physical storage and the student is notified. We can see that this process requires many clearances and manual confirmations. This requires a lot of time and manpower. They also maintain a student application file for all the students which uses a lot of paper.

Fig. 2. As-is model of departmental archiving

In Fig. 2, the departmental archiving uses a huge amount of paper as they photocopy and archive course files, mid and final scripts, and departmental documents. This is also time and money-consuming. From the interviews and the author’s idea, the following to-be model is proposed. Figure 3 shows that almost all the processes can be conducted online from submitting documents to notifying different departments like the Registrar’s Office, Library, Accounts, and Exam office. Without receiving paper applications physically, the university can accept these documents through the East West University online student portal. The files can get clearances one after the other without the need to manual handling and send the paper documents. A graduation application file consists of 3 papers for a student. i.e. main application, course checklist, and grade report. According to the interviewees, storing grade reports online is risky since they fear data center failure and hence strongly suggested physical storage of the grade reports which means 2 pages per student can be still saved with the proposed model. In Fig. 4, the authors proposed scanning files rather than photocopying. After scanning it can be stored in cloud storage or the university’s server.

374

R. M. Haque et al.

Fig. 3. To-be model of Graduation application process.

Fig. 4. To-be model of departmental archiving

If only course file archiving is considered, each semester 9 exam scripts (each consisting of 8 pages on average), 3 exam questions, 1 outline, 1 grade list, and an average of 5 lab manuals are archived for each course. Table 1 shows the approximate pages required for the Graduation Application process. 3 pages per student are required for this process. This number summed up to 5079 pages for all the students in the year 2022.

Drivers and Barriers for Going Paperless in Tertiary

375

Table 1. Paper usage of the graduation application process

Graduation application

For 1 student (pages)

For all students per year (pages)

3

5079 (year 2022)

Table 2. Paper usage of course archiving process

Course archiving

For 1 course (pages)

For all courses per year (pages)

82

29,192 (year 2019)

Table 2 shows the approximate pages required for the Course Archiving process. 82 pages per course is required for this process. This summed up to 29,192 pages for all courses in the year 2019. Paper usage of the last 5 years for both processes is given below.

Fig. 5. Paper usage of last 5 years of giving two processes.

Figure 5 summarizes the paper usage of the 2 processes between the years 2017 and 2021. The blue and orange colored plot shows the paper usage in the last 5 years of the Graduation Application and the Departmental Archiving process respectively.

5 Discussion Previous works do not present any estimation of paper usage, particularly on graduation applications and departmental archiving processes. Moreover, though some mentioned certain ways to turn different processes paperless, no as-is or to-be models of the processes were shown.

376

R. M. Haque et al.

From the interviews taken, authors got to know that, as of today no specific paperless initiatives have been taken. University authority is discussing turning the graduation application process online, but since the student portal is still developing the initiative has yet to be taken. The results from this paper can be used as a guide to estimate the paper usage of the mentioned processes and the two to-be models proposed in this paper can be used if needed. Figure 5 shows the amount of paper used between the years 2017 and 2019 for the two processes: The graduation Application process and Departmental Archiving. However, in the years 2020 and 2021, these processes took place online for Covid-19, and hence the figure shows 0 paper usage for departmental archiving. However, the graduation application process still required storing a single copy of the grade report of each student manually during these two years which is represented by the reduced but still visible amount of paper used by this process in the graph. All these thousands of papers could be saved by implementing the to-be model. The 5 interviewees showed great enthusiasm to turn these two processes paperless. However, some barriers were also found. The authors have tried to summarize the main drivers and barriers to going paperless in the following two tables: Table 3. Drivers for going paperless Graduation application

Departmental archiving

Maintenance of student files needless

Storing paper documents needless, reduces physical storage space

Ensuring several clearances on paper is troublesome

Faster and easier sorting and searching of files according to dates, course, and section

Manual checking course checklists is needless, saves time and manpower

Disposing of documents like mid and final scripts is needless

Cost reduction as less number of papers will be needed

Cost reduction as less number of papers will be needed

Table 3 summarizes the main drivers for a university wanting to go paperless. The first and second column shows the drivers for the Graduation Application and Departmental Archiving process respectively. Table 4 summarizes the main barriers that stand in the way of the university making the processes paperless. The first and second column shows the barriers to the Graduation Application and Departmental Archiving process respectively.

Drivers and Barriers for Going Paperless in Tertiary

377

Table 4. Barriers to going paperless Graduation application

Departmental archiving

Course checklist is considered to be a very To archive the mid and final scripts, they must critical section which is risky to leave it to the be scanned first which needs manpower IT section only Grade reports encouraged to be saved physically fearing data center failure

Officials who come for accreditation purposes prefer paper-based files to review than scanned/soft copies of the paper

6 Conclusion The objective of this study was to find the drivers and barriers to going paperless in education institutes. The authors found that no specific paperless initiative has been taken by the university yet. The author described that by changing the existing process model university can reduce paper usage, reduce manpower, save time, and also save physical space. Due to the limitation of time and resources, the authors considered only the CSE department for the departmental process and only one administrative process of East West University. For less time authors couldn’t conduct the study at other universities. As a continuation of this study, different departments and processes of different universities could be explored since paper usage will vary among different departments, processes, and universities. As this study was based on a private university, so same study on a public university might tell some other results. This then can also be used to make a comparison between different departments and universities.

References 1. Kayode, A.A., Lawan, B.M., Ajadi, I.A., Lukman, J.A.: E-government, information and communications technology support and paperless environment in Nigerian public universities: issues and challenges. J. Technol. Manag. Bus. 7 (2020). http://doi.org/10.30880/jtmb.2020. 07.01.006 2. Khan, R.A., Al Mesfer, M.K., Khan, A.R., Khan, S., Van Zutphen, A.: Green examination: integration of technology for sustainability. Environ. Dev. Sustain. 19(1), 339–346 (2015). https://doi.org/10.1007/s10668-015-9736-9 3. Isaeva, M., Yoon, H.Y.: Paperless university—how we can make it work? In: 2016 15th International Conference on Information Technology Based Higher Education and Training, ITHET 2016 (2016) 4. Prastyo, P.H., Sumi, A.S., Kusumawardani, S.S.: A systematic literature review of application development to realize paperless application in Indonesia: sectors, platforms, impacts, and challenges. Indonesian J. Inf. Syst. 2, 111–129 (2020). http://doi.org/10.24002/ijis.v2i2.3168 5. Srimathi, H., Krishnamoorthy, A.: Paperless administration in Indian higher education. Int. J. Eng. Adv. Technol. 8, 760–764 (2019) 6. Hafiz Iqbal, M.: Paperless campus: the real contribution towards a sustainable low carbon society (2015). https://doi.org/10.9790/2402-09811017

378

R. M. Haque et al.

7. Meishar-Tal, H., Shonfeld, M.: Students’ writing and reading preferences in a paperless classroom. Interact. Learn. Environ. 27 (2019). https://doi.org/10.1080/10494820.2018.150 4306 8. Genesis, E., Oluwole, O.: Towards a “paperless” higher education system in Nigeria: concept, challenges and prospects. J. Educ. Soc. Behav. Sci. 24 (2018). https://doi.org/10.9734/jesbs/ 2018/19913 9. Baby, K.T., Saeed, M.A.: Beyond the classroom through the paperless mode. Int. J. Linguist. Lit. Transl. 2 (2020). http://doi.org/10.32996/ijllt.2020.3.1.9 10. Sofia, L.M., Umaima, F.F., Rumyaru, B.: Going paperless concept implementation at senior high school in Bekasi, Indonesia. J. Environ. Eng. Waste Manag. 6 (2021). https://doi.org/10. 33021/jenv.v6i1.1344 11. Ochi, K.: Students’ perception of paperless English classroom: a case study of a Japanese it university campus. Teach. English Technol. 21, 35–50 (2021) 12. Galanis, P.: Methods of data collection in qualitative research. Arch. Hell. Med. 35 (2018) 13. Peltier, T.R.: Gap analysis. Inf. Secur. Risk Anal. 116–127 (2021). https://doi.org/10.1201/ ebk1439839560-9

Impact of Lifestyle on Career: A Review Md. Jabed Hosen1 , Md. Injamul Haque1 , Saiful Islam1 , Mohammed Nadir Bin Ali2 , Touhid Bhuiyan1 , Ahmed Wasif Reza3(B) , and Mohammad Shamsul Arefin1,4(B) 1 Department of Computer Science and Engineering, Daffodil International University, Dhaka,

Bangladesh {hosen15-3834,injamul15-3798,saiful15-3809}@diu.edu.bd, [email protected] 2 Department of Tourism and Hospitality Management, Daffodil International University, Dhaka, Bangladesh [email protected] 3 Department of Computer Science and Engineering, East West University, Dhaka, Bangladesh [email protected] 4 Department of Computer Science and Engineering, Chittagong University of Engineering and Technology, Chattogram 4349, Bangladesh

Abstract. In recent years, the impact of lifestyle on career has grown in significance. The way a person lives has a significant impact on different aspects of life and this has been a topic of interest among many scholars and practitioners alike. The type of lifestyle a person considers, as well as their potential for success in their career, can be influenced by their lifestyle choices. This article has reviewed more than fifty papers based on various lifestyle choices. The review emphasizes the relationship between lifestyle choices and career outcomes, implying that living a healthy lifestyle can lead to increased work productivity and career success. This study also aims to shed light on the major lifestyle factors and how these factors influence various aspects of life such as work-ability, mental and physical health, recreation, travel, sleep, smoking, diet, and life. Keywords: Career · Lifestyle · Work-ability · Life expectancy · Health · Mental health · Physical health · Recreation · Travel · Sleep · Smoking · Diet

1 Introduction Lifestyle refers to the way people live, including the place they live in with their family and society, the job they do, and the activities they enjoy. Lifestyle has an impact on almost every aspect of life. The way in which a person lives reveals a lot of things about that person such as physical health, mental health, life expectancy, productivity, and other variables of life. Lifestyle describes the traits of people who reside in a particular time and area. It comprises people’s regular activities and behaviors related to their work, hobbies, and food [1]. It is also commonly seen that people who have a mental illness also have poor physical health, and in comparison to the general population they have significantly higher rates of different physical disorders [2–4]. When examining another © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 379–393, 2024. https://doi.org/10.1007/978-3-031-50158-6_38

380

Md. Jabed Hosen et al.

aspect of lifestyle it is found that there are a number of unhealthy habits, attitudes and behaviors which can lead to chronic disease and mortality. Lack of physical exercise, tobacco use, poor diet [5–9], high cholesterol levels [10, 11], and inadequate dental care appears to be significant factors in poor physical health [12, 13]. Understanding the effects of different lifestyle factors allows us to make more informed decisions about our own behaviors and work toward promoting healthier habits that can improve people’s overall health and well-being. The paper will also consider how larger societal factors can influence individual lifestyle choices and community health outcomes. Overall, the goal of this article is to provide insights into the relationship between lifestyle and various aspects of life, emphasizing the importance of making informed behavioral choices and promoting healthier lifestyles to improve our overall well-being [14]. Healthy life and good productivity can lead a person to a better career. So, it will be possible to draw a relationship between lifestyle and career.

2 Methodology This study summarizes the available research on lifestyle variables that contribute to poor physical health, and mental health, including low concentrations of exercise, poor diet and nutrition, elevated cholesterol levels, smoking use, and recreation all contribute to a higher risk for cardiovascular disease and productivity. A multidisciplinary search of online databases and journals was done to create an integrative review, with a focus on the career, health, and lifestyle issues that were most frequently discussed in the literature. 2.1 Phase 1-Planning This section describes how the relevant papers were chosen. The papers were collected from prestigious sources such as Springer, SAGE, Science Direct, BMC, and MDPI, as well as PubMed. This audit described the following search terms: “Impact of Lifestyle on Physical Health,” “Impact of Lifestyle on Mental Health,” “Impact of Lifestyle on Work-ability,” “Impact of Lifestyle on Recreation,” “Impact of Lifestyle on Expectancy,” and “Impact of Lifestyle on Sleep”. 2.2 Phase 2-Conducting In this phase, all articles are rigorously checked for reliability and validity in order to be accepted as the final sample article for review. 2.3 Phase 3-Reporting After careful consideration, 51 relevant research papers were selected for review. Then we tried to find a relationship between the career and those six factors from the review papers. The categorized evaluation identifies contributions to research, work processes, and flaws. As a result, all papers are carefully selected to meet the goal of this research (Fig. 1).

Impact of Lifestyle on Career: A Review

381

Fig. 1. Shows the six lifestyle factors.

3 Paper Collection The main purpose of this section is to describe how we collected papers for our research. Despite a lack of available papers on recreation, we were able to find research dating back to 1956. Then we started reviewing papers that were published in 1956 and other papers that were published in 1996. Unfortunately, it was difficult to find papers in 1996, 1997, 2001, 2004, 2007, 2009, 2011, 2013, 2019 and 2021 individually. So, we merged these years as there were few papers found in these years. In the end, we managed to gather 51 papers which fulfill the requirements of six categories mentioned in Sect. 2. Figure 2: There are many papers published on Physical Health and Mental Health, but there are few articles on Work-ability, Recreation, Life Expectancy, and Sleep, among the six categories mentioned above. The distribution of categories

Fig. 2. Shows the distribution of selected papers by category.

382

Md. Jabed Hosen et al.

Table 1 represents our selected papers based on publishing time scale and above mentioned six categories.

4 Detailed Review of Papers Impact of Lifestyle on Career Lifestyle has a relation to physical health, mental health and productivity. And these variables which are partially or fully dependable on lifestyle have a connection to a career. These variables are: 4.1 Impact of Lifestyle on Physical Health Living a healthy lifestyle and being healthy are related. People having chronic diseases have some prevalent lifestyle traits, including smoking, eating unhealthy, being inactive, and also being obese [15]. The main causes of mortality in the United States are tobacco use, a poor diet, and a lack of physical activity [16]. The inclusion of these lifestyle factors in important diet-related health reports can help establish their significance for public health [17]. The change of lifestyle is emphasized as a crucial component of prevention and control in treatment guidelines regarding blood pressure [21], cholesterol [22], and obesity [23], as well as with regard to smoking [18], physical activity [19], and blood pressure [20]. Among adult individuals, there is a concentration of risky lifestyle behaviors such as smoking, insufficient consumption of fruits and vegetables, drinking too much alcohol, and not doing enough physical exercise [24]. A process that alters lifestyles and gives people more control over their health is personal health or healthy lifestyle practices. Based on the decisions you make about your daily routines, leading a healthy lifestyle makes you fit, energetic, and at a lower risk for disease. A balanced diet, consistent exercise, and sufficient sleep are the cornerstones of keeping one’s health. According to research, employees who are healthy have the highest chance of succeeding [25]. Ozvurmaz S, Mandiracioglu A, et al. They said a workplace has a direct impact on employees’ physical, emotional, economic, and social wellness [26]. In addition, he said that 2 million individuals worldwide pass away every year as a result of workplace diseases and accidents. A healthy diet, physical activity, weight control, and stress reduction are just a few of the strategies to live a long and healthy life. Additionally, some data suggests that in some situations, higher levels of well-being may have a direct effect on levels of work performance (Table 2). 4.2 Impact of Lifestyle on Mental Health According to the hypotheses that guide the science of social epidemiology, contemporary society has a persistent negative impact on public health [27]. Jonsdottir et al.; Xu et al. In addition to characteristic indicators, everyday circumstances, and big life events, rising research suggests that everyday actions that can be changed by a person can have an impact on mental health. Prospective studies commonly find a symbiotic relationship between many lifestyle factors and both mental and physical wellness, with notable health benefits and well-being happening in reply to very little changes in lifestyle [28,

1996–2000

Leisure and lifestyle [40] 1959 2001–2005

Oral health [12]

2010–2023

Prospective study [28]

Somatic healthcare [13]

Meta-analytic [39]

Physical exercise and health [34]

Health [33]

Lifestyle factors [43]

Lifestyle profiling work [41] 1979 2006–2010

Global health [47]

Lifestyle risk factors [44]

(continued)

Population based study [46]

Life expectancies [42]

2010–2023

Physical activity [38]

Mental illness [7]

Mental health problems [8]

Nutrition and mental health [9]

Mental health programs [32]

Career [37]

Comorbid mental illness[4]

Public health [31]

Mental health [28]

Systematic review [36]

Depressive disorder [10]

Psychosocial health [26]

Metabolic syndrome[3]

2010–2023

Public health [1]

Healthy lifestyles [26]

Lifestyles and job performance [25]

Lifestyle index [35]

Health and lifestyle [6]

2006–2010

2006–2010 Psychiatric patients [11] Modern health [14]

Surveillance system [30]

2001–2005

Causes of death [16]

High blood pressure [21]

Adult population [24] Adult treatment panel [22]

Physical illness [2]

Chronic disease [15]

1996–2000

Life expectancy based research

Recreation based research

Work-ability based research

Mental health based research

2001–2005 Healthy lifestyle [17] Women and smoking [19]

Reducing tobacco [18] Bipolar disorder [5]

Obesity in adults [23]

Physical health based research Physical activity [20]

Table 1. Evolutions of impact of lifestyle on career

Impact of Lifestyle on Career: A Review 383

Sleep based research

Prospective study [51]

2001–2005 Systematic review [49]

Sleep loss [50]

2006–2010

Table 1. (continued)

Life expectancy [45]

2010–2023 Healthy lifestyle [48]

384 Md. Jabed Hosen et al.

Impact of Lifestyle on Career: A Review

385

Table 2. Physical health based research Paper title

Contribution

Healthy lifestyle [17] American adults’ propensity for healthy lifestyle behaviors Reducing tobacco [18]

Dataset

Evaluation

National data for the year 2000 (telephone surveys)

Research has been promoted

Effective strategies Did not use a specific to reduce tobacco use dataset

Women and smoking Raise awareness [19] about the harmful effects of smoking among women

Provided a strong scientific basis for policy interventions to control tobacco use

Comprehensive Targeted efforts to analysis of the reduce smoking literature and research among women on the health impacts of smoking on women

Physical activity [20] Increasing public Did not use a specific knowledge of the dataset for analysis value of exercise for health and wellbeing

Promoting public health initiatives to increase physical activity levels

Modern health [14]

Promotion of modern health and well-being

Sustainable approach Does not provide any to modern health care original data

29]. Brown DW et al. their studies state that the metrics used in demographic studies to assess both mental and physical well-being have evolved over time. While considered essential for patient studies in medical contexts, lengthy equipment is impractical for observational studies. In population studies that have been written up in the international literature, one item has come to be the standard for evaluating overall health [27]. Although cross-sectional research like this one cannot be used to draw conclusions about causality, the notion that lifestyle will affect self-rated mental health is well established. Rarely are results related to mental health included in evaluations of public health initiatives intended to promote behaviors [30]. Based on a systematic review of published research on academic well-being, academics are more likely to have mental health problems than people in other professions. According to research conducted by Hsiao et al. [31], factors affecting academics’ health include a lack of employment security, limited management support, and the responsibility of job demands on their time. Uedo and Niino, et al.The authors of research on the impact of mental health programs on worker productivity discovered that there is a statistically significant link between health and productivity and that the performance of the firms that offer more highly regarded health program practices is higher [32]. Similar to vein, Yu and Bang [33] explore the effects of better health on organizational performance. The findings showed that employees with poor well-being are significantly more likely than employees with high well-being to engage in behaviors that would have a detrimental impact on organizational outputs, both in regard to direct healthcare expenses and organizational evaluation methods [33].

386

Md. Jabed Hosen et al.

Drannan [34] carries out research on the connection between physical activity and work performance; the findings showed that there was a strong connection. The typical justification for starting and maintaining a regular exercise regimen is the health advantages of physical activity. Research has shown that exercise significantly enhances both mood and productivity. Numerous psychologists and leading businesses have included physical activity in their eight corporate strategies to improve employee productivity by boosting mood and job performance [34] (Table 3). Table 3. Mental health based research Paper title

Contribution

Dataset

Evaluation

Prospective study [2] Leisure - time physical activity (PA) and depression and anxiety symptoms, burnout, and perceived stress

Data from a cohort of health and social insurance employees in west Sweden between 2004 and 2006

Able to find association between physical activity and different types of anxiety

Somatic healthcare [13]

Research on the use of community mental health services by people with severe mental illness

200 dataset where 100 with schizophrenia and 100 with affective disorder

The general population to report having sought some form of medical care in the past year

Health and lifestyle [6]

Design and implement Survey data a structured educational program

Improving the health and mental health services and promoting healthy behavioral changes

Surveillance system [30]

Encouraging physical exercise to enhance overall quality of life related to health

(BRFSS) surveyData

Exercise improves health-related quality of life

Public health [31]

Positive impact on the students’ knowledge

Dataset collected from The development of the nursing students effective health promotion interventions for nursing students

4.3 Impact of Lifestyle on Work-Ability DOROTA KALETA et al. state that lifestyle has a direct impact on workability [35]. Tilja van den Berg et al. said that certain personal traits and work lifestyle requirements are linked to workability. And they state that some factors that are associated with decreased workability are older age, obesity, absence of intense exercise during free time, and many more [36]. Hussein Isse Hassan Abdirahman et al. developed three hypotheses and the

Impact of Lifestyle on Career: A Review

387

outcome of those hypotheses states that Job balancing and productivity are positively correlated [37]. Also, sometimes it affects students’ college life. JESSE CALESTINE et al. discovered that, compared to younger age groups, college students may have a distinct link between academic results and physical exercise. This study sheds light on the creation of upcoming university college health interventions that will jointly prioritize academic results and physical exercise [48]. Mesmer-Magnus, J.R., Viswesvaran, C. et al. Their review has discovered that parental involvement for company culture helps with work-life balance management. Employee motivation is increased, and the negative effects of work-life balance are lessened when there is support from superiors, coworkers, and flexible working hours. Proper maternity leave, etc. Conflicts between work and life were further exacerbated by this [39] (Table 4). Table 4. Work-ability based research Paper title

Contribution

Dataset

Lifestyle index [35]

Lifestyle factors in Datasets were used maintaining good work in this project: ability among interview employees

Employees with a healthier lifestyle had higher work ability

Career [37]

Organizational behavior and human resource management

Data collected from 184 employees using a survey questionnaire

Job satisfaction and enhancing employee performance

Meta-analytic [39]

How family-friendly workplaces can reduce conflict between work and family

30 research studies that looked into this meta-analysis

Correlated with reduced levels of conflict between work and family

Physical activity [38]

Association between a Collected primary college student’s work data through surveys habits and their level of and questionnaires fitness and physical activity

College students who engage in regular physical activity have better work habits

Systematic review [36]

Identifies important personal and professional aspects that affect job performance

Promote healthy and productive work environments

Using various databases such as PubMed, Embase, and PsycINFO

Evaluation

4.4 Impact of Lifestyle on Recreation ROBERT J. HAVIGHURST et al. states that There are two main categories of leisure style: one that is focused on the public as well as one that is focused on the home [40]. Both a society living and a residence lifestyle include leisure activities that are focused on

388

Md. Jabed Hosen et al.

the local area. Based on the interview questionnaire they gave scores to eight social roles. From the score of the eight social roles, they divided leisure time into home-centered and community-centered. The relationship between a person’s lifestyle and their leisure activities is not always close. Approximately 5% of adults fall into one of these categories. They are people who live prosperous lives and have good personal adjustment but do little to no leisure activity. These men and women typically devote the majority of their energy to their jobs or to their homes and children, with little leisure time or desire. Nearly 6% of adults belong to a different group. They have a lot of hobbies but struggle to be successful as workers, parents, or husbands because they feel inadequate in these roles. Traveling is another popular recreational activity. JAMES R. ABBEY discovered that tourists prefer tours created using vacation lifestyle information over those created using de demographics (Age, Gender, Relationship Status, Degree, Profession, Income, Family Make-up, Prior Travel Experience) data, and this demographic preference is consistent for both different trip types (air and motor-coach) and different priced tours (budget and first-class) [41]. So, it is clear that the selection of a travel package depends a lot on the lifestyle of individuals (Table 5). Table 5. Recreation based research Paper title

Contribution

Leisure and lifestyle [40]

Understanding of the Interview role of leisure in shaping questionnaire lifestyles and social identities

Dataset

Framework for understanding the relationship between leisure and lifestyle

Lifestyle profiling work [41]

Understanding of lifestyle profiling as a marketing tool in the travel industry

Lifestyle profiling could effectively predict travel

Questionnaire

Evaluation

4.5 Impact of Lifestyle on Expectancy Their studies state that good lifestyle decisions, such as maintaining a balanced diet, exercising frequently, and giving up smoking and drinking too much alcohol, have a positive effect. on health, particularly in terms of increased lifespan [42–44]. Mehta N et al. in their review, it is pertinent to inquire as to if smokers who give up early in life are still likely to enjoy long, ailment-free lives as the largest smoking group in their sample was former smokers (41 Percentage of the overall population). Separate analyses (data not shown) revealed that non-obese individuals who had quit smoking at least 10 years before the survey and who drank moderately had lifespans that were only one year less than non-obese people who had not previously smoked and who drank moderately [45]. Rizzuto D et al. state that, Smokers had a year poorer survival rate than non-smokers among those who survived to just be 75 years old. In the Kungsholmen

Impact of Lifestyle on Career: A Review

389

Sample population, 83% of ex-smokers had quit 15–35 years earlier than baseline, and 17% had quit five to 14 years earlier [46]. Li Y et al. According to their calculations, adopting 5 low-risk lifestyle-related traits may increase a person’s average lifespan at the age of 50 by 14.0 and 12.2 years, respectively, for male and female US citizens [42]. As the life expectancy estimates are simpler to comprehend by both the whole public as well as medical specialists, they have grown in popularity as a statistic for establishing public health objectives. The relationship between single and combination lifestyle choices including eating, drinking, and tobacco and expected lifespan has not yet been studied with regard to the occurrence of multimorbidity [47, 48]. Rizzuto D et al. Among individuals with one or more chronic conditions, only one study examined the connection between a combined healthy lifestyle and life expectancy [46], Chudasama YV et al. While their studies involved people from the general community, the results revealed that a healthy lifestyle overall was linked to an average lifespan of between 5.4 and 18.9 years [48] (Table 6). Table 6. Life expectancy based research Paper title

Contribution

Dataset

Evaluation

Lifestyle factors [43]

Significant factor affecting life expectancy in the Japanese population

Over 100,000 participants aged 40–79 years the baseline survey collected data on various lifestyle factors

Alcohol consumption and physical activity had a more modest impact on life expectancy

Lifestyle risk factors [44]

Living a healthy lifestyle has a negative impact on Germans’ remaining life expectancy

Over 10,000 participants aged 40–79 years for a period of 10 years

Healthy lifestyles and reducing the burden of non-communicable diseases

Life expectancy [45]

Healthy lifestyle can lead to significant population health benefits

Data from the (HRS), a survey

Healthy lifestyle participants lived 7.6 years longer

Global health [47]

Multimorbidity is associated with increased healthcare utilization

Paper draws on a Effective models of range of sources, care for individuals including with multimorbidity published research studies, reports, and policy documents

390

Md. Jabed Hosen et al.

4.6 Impact of Lifestyle on Sleep Shneerson J et al. According to this study, there isn’t enough data from randomized controlled trials to determine how well weight loss, exercise, and sleep hygiene approaches work to cure obstructive sleep apnea. This is particularly significant because this prevalent ailment is one for which these procedures are frequently advised [49]. Akerstedt T. et al. state that, in addition to these variations in compensation calculations, there are also variations in how health status is viewed. In certain American jurisdictions, doctors are required to disclose whether a professional driver has a sleep issue. Such a reporting requirement might materially affect the driver’s ability to maintain their employment [50]. In contrast to the transportation industry, the statistics of sleep-related incidents in industrial and health. One explanation might be because, in contrast to the transportation industry, where errors can have catastrophic consequences, there are not as many jobs in these industries that include a lot of duties with a significant injury risk [41]. However, it was shown in epidemiological research by Akerstedt et al. that shift work and disrupted sleep had a 50% database with a larger danger of fatal workplace accidents of more than 50,000 people who had been randomly recruited from the community [51] (Table 7). Table 7. Sleep based research Paper title

Contribution

Dataset

Evaluation

Systematic review [49]

Effectiveness of lifestyle modifications in the treatment of OSA

Reviewed 13 randomized controlled trials that evaluated various lifestyle modifications

Effectiveness of other lifestyle modifications is less clear

Sleep loss [50]

Research of how sleep loss affects several facets of human ability

Information from multiple studies and sources

Lack of sleep has a detrimental impact on performance and raises the possibility of accidents

Prospective study [51]

The role of sleeping difficulties and occupational factors in fatal occupational accidents

A prospective study of 121,390 male workers in Sweden

The prevalence of sleeping difficulties and improving working conditions

5 Discussion After reviewing these 51 studies, it’s clear that lifestyle has a direct or indirect impact on other aspects of life. Physical, mental and life expectancy are all highly dependent on lifestyle. Common variables that were analyzed to find the relationship between lifestyle and health are smoking, alcohol consumption level, sleeping time, stress level, and diet.

Impact of Lifestyle on Career: A Review

391

A healthy person has a better life expectancy. With a healthy mind and healthy body, a person can have better work efficiency. While recreation is important to keep good mental health, lifestyle has a significant impact on the choice of recreational activity as well. To live a healthy life with a healthy mind and have better work efficiency a person needs to have a healthy lifestyle.

6 Conclusion This literature review provides useful insights into the impact of lifestyle on various aspects of life. It emphasizes the importance of adopting healthy lifestyle practices like frequent exercise and a balanced diet, and adequate sleep in promoting better physical and mental health outcomes. We did a thorough analysis of the literature on human behavior and lifestyle, focusing on 51 publications and case studies that were published between 1959 and December 2023, in order to accomplish our aim. We finally chose 51 main papers for our study out of the 80 initially chosen publications using various inclusion and exclusion criteria. We did a thorough assessment of all of the research articles. The analysis of these research papers discovered that people who are physically active throughout the day and in generally good spirits have a healthy life as a result of healthy lifestyles. Overall, people with a healthy body give consideration to getting enough sleep and eating a balanced diet. And a person’s lifestyle can have a significant impact on their career. People who have a better career live a healthier lifestyle, which allows them to work more efficiently.

References 1. Farhud, D.D.: Impact of lifestyle on health. Iran. J. Public Health 44(11), 1442 (2015) 2. Holman, D.: Duty to Care, Preventable Physical Illness in People with Mental Illness. Department of Public Health, University of Western Australia (2001) 3. John, A.P., Koloth, R., Dragovic, M., Lim, S.C.: Prevalence of metabolic syndrome among Australians with severe mental illness. Med. J. Aust. 190(4), 176–179 (2009) 4. Mitchell, A.J., Malone, D., Doebbeling, C.C.: Quality of medical care for people with and without comorbid mental illness and substance misuse: systematic review of comparative studies. Br. J. Psychiatry 194(6), 491–499 (2009) 5. Elmslie, J.L., Mann, J.I., Silverstone, J.T., Romans, S.E.: Determinants of overweight and obesity in patients with bipolar disorder. J. Clin. Psychiatry 62(6), 4297 (2001) 6. O’sullivan, J., Gilbert, J., Ward, W.: Addressing the health and lifestyle issues of people with a mental illness: the healthy living programme. Australas. Psychiatry 14(2), 150–155 (2006) 7. Osborn, D.P., Nazareth, I., King, M.B.: Physical activity, dietary habits and coronary heart disease risk factor knowledge amongst people with severe mental illness: a cross sectional comparative study in primary care. Soc. Psychiatry Psychiatr. Epidemiol. 42, 787–793 (2007) 8. Soundy, A., Faulkner, G., Taylor, A.: Exploring variability and perceptions of lifestyle physical activity among individuals with severe and enduring mental health problems: a qualitative study. J. Ment. Health 16(4), 493–503 (2007) 9. Porter, J., Evans, S.: Nutrition and mental health research in Australia and New Zealand: a review of progress and directions for the future. Nutr. Diet. 65(1), 6–9 (2008) 10. Jow, G.M., Yang, T.T., Chen, C.L.: Leptin and cholesterol levels are low in major depressive disorder, but high in schizophrenia. J. Affect. Disord. 90(1), 21–27 (2006)

392

Md. Jabed Hosen et al.

11. Ojala, K., Repo-Tiihonen, E., Tiihonen, J., Niskanen, L.: Statins are effective in treating dyslipidemia among psychiatric patients using second-generation antipsychotic agents. J. Psychopharmacol. 22(1), 33–38 (2008) 12. Sjögren, R., Nordström, G.: Oral health status of psychiatric patients. J. Clin. Nurs. 9(4), 632–638 (2000) 13. Dickerson, F.B., McNary, S.W., Brown, C.H., Kreyenbuhl, J., Goldberg, R.W., Dixon, L.B.: Somatic healthcare utilization among adults with serious mental illness who are receiving community psychiatric services. Medical Care 560–570 (2003) 14. Warbrick, I., Makiha, R., Heke, D., Hikuroa, D., Awatere, S., Smith, V.: Te Maramataka— an indigenous system of attuning with the environment, and its role in modern health and well-being. Int. J. Environ. Res. Public Health 20(3), 2739 (2023) 15. Brownson, R.C., Remington, P.L., Davis, J.R.: Chronic Disease. Epidemiology and Control. American Public Health Association, Washington, DC (1998) 16. Mokdad, A.H., Marks, J.S., Stroup, D.F., Gerberding, J.L.: Actual causes of death in the United States, 2000. JAMA 291(10), 1238–1245 (2004) 17. Reeves, M.J., Rafferty, A.P.: Healthy lifestyle characteristics among adults in the United States, 2000. Arch. Intern. Med. 165(8), 854–857 (2005) 18. US Department of Health and Human Services: Reducing tobacco use: a report of the Surgeon General (2000) 19. Office on Smoking and Health: Women and smoking: a report of the Surgeon General (2001) 20. National Center for Chronic Disease Prevention, Health Promotion (US), President’s Council on Physical Fitness and Sports (US): Physical activity and health: a report of the Surgeon General. US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion (1996) 21. Chobanian, A.V., et al.: The seventh report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure: the JNC 7 report. JAMA 289(19), 2560–2571 (2003) 22. National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) final report. Circulation 106(25), 3143–3421. PMID: 12485966 23. Expert Panel on the Identification, Evaluation, Treatment of Overweight, Obesity in Adults (US), National Heart, Lung, Blood Institute, National Institute of Diabetes, Digestive and Kidney Diseases (US) (1998) Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: the evidence report (No. 98). National Institutes of Health, National Heart, Lung, and Blood Institute 24. Schuit, A.J., van Loon, A.J.M., Tijhuis, M., Ocké, M.C.: Clustering of lifestyle risk factors in a general adult population. Prev. Med. 35(3), 219–224 (2002) 25. Agboola, I.O., Ikonne, C.N.: Healthy lifestyles and job performance of academics: a theoretical perspectives. Libr. Philos. Pract. (2019) 26. Ozvurmaz, S., Mandiracioglu, A.: Healthy lifestyle behavior of employees in small and medium-sized enterprises in Aydin, Turkey. Pak. J. Med. Sci. 33(2), 404 (2017) 27. Pikhart, H., Bobak, M., Rose, R., Marmot, M.: Household item ownership and self-rated health: material and psychosocial explanations. BMC Public Health 3(1), 1–7 (2003) 28. Jonsdottir, I.H., Rödjer, L., Hadzibajramovic, E., Börjesson, M., Ahlborg, G., Jr.: A prospective study of leisure-time physical activity and mental health in Swedish health care workers and social insurance officers. Prev. Med. 51(5), 373–377 (2010) 29. Xu, Q., Anderson, D., Courtney, M.: A longitudinal study of the relationship between lifestyle and mental health among midlife and older women in Australia: findings from the healthy aging of women study. Health Care Women Int. 31(12), 1082–1096 (2010)

Impact of Lifestyle on Career: A Review

393

30. Brown, D.W., et al.: Associations between recommended levels of physical activity and healthrelated quality of life findings from the 2001 Behavioral Risk Factor Surveillance System (BRFSS) survey. Prev. Med. 37(5), 520–528 (2003) 31. Hsiao, Y.C., Chen, M.Y., Gau, Y.M., Hung, L.L., Chang, S.H., Tsai, H.M.: Short-term effects of a health promotion course for Taiwanese nursing students. Public Health Nurs. 22(1), 74–81 (2005) 32. Ueda, Y., Niino, H.: The effect of mental health programs on employee satisfaction with benefit programs, jobs and the organization. Bus. Manag. Rev. 2(1), 27–38 (2012) 33. Yu, K., Bang, S.C.: What is the Impact of Improved Health to Organizational Performance? (2013) 34. Drannan, J.D.: The relationship between physical exercise and job performance: the mediating effects of subjective health and good mood. Doctoral dissertation, Bangkok University (2016) 35. Kaleta, D., Makowiec-Dabrowska, T., Jegier, A.: Lifestyle index and work ability. Int. J. Occup. Med. Environ. Health 19(3), 170 (2006) 36. Van den Berg, T., Elders, L., de Zwart, B., Burdorf, A.: The effects of work-related and individual factors on the work ability index: a systematic review. Occup. Environ. Med. 66, 211–220 (2008) 37. Abdirahman, H.I.H.: The relationship between job satisfaction, work-life balance and organizational commitment on employee performance (2018) 38. Calestine, J., Bopp, M., Bopp, C.M., Papalia, Z.: College student work habits are related to physical activity and fitness. Int. J. Exerc. Sci. 10(7), 1009 (2017) 39. Mesmer-Magnus, J.R., Viswesvaran, C.: How family-friendly work environments affect work/family conflict: a meta-analytic examination. J. Lab. Res. 27(4), 555–574 (2006) 40. Havighurst, R.J., Feigenbaum, K.: Leisure and life-style. Am. J. Sociol. 64(4), 396–404 (1959) 41. Abbey, J.R.: Does life-style profiling work? J. Travel Res. 18(1), 8–14 (1979) 42. Li, Y., et al.: Impact of healthy lifestyle factors on life expectancies in the US population. Circulation 138(4), 345–355 (2018) 43. Tamakoshi, A., et al.: Impact of smoking and other lifestyle factors on life expectancy among Japanese: findings from the Japan Collaborative Cohort (JACC) study. J. Epidemiol. 20(5), 370–376 (2010) 44. Li, K., Hüsing, A., Kaaks, R.: Lifestyle risk factors and residual life expectancy at age 40: a German cohort study. BMC Med. 12(1), 1–10 (2014) 45. Mehta, N., Myrskylä, M.: The population health benefits of a healthy lifestyle: life expectancy increased and onset of disability delayed. Health Aff. 36(8), 1495–1502 (2017) 46. Rizzuto, D., Orsini, N., Qiu, C., Wang, H.X., Fratiglioni, L.: Lifestyle, social factors, and survival after age 75: population based study. Bmj 345 (2012) 47. Academy of Medical Sciences (Royaume uni): Multimorbidity: a priority for global health research. Academy of Medical Sciences (2018) 48. Chudasama, Y.V., et al.: Healthy lifestyle and life expectancy in people with multimorbidity in the UK Biobank: a longitudinal cohort study. PLoS Med. 17(9), e1003332 (2020) 49. Shneerson, J., Wright, J.J., Cochrane Airways Group: Lifestyle modification for obstructive sleep apnoea. Cochrane Database Syst. Rev. 2010(1) (1996) 50. Åkerstedt, T., Philip, P., Capelli, A., Kecklund, G.: Sleep loss and accidents—work hours, lifestyle, and sleep pathology. Prog. Brain Res. 190, 169–188 (2011) 51. Åkerstedt, T., Fredlund, P., Gillberg, M., Jansson, B.: A prospective study of fatal occupational accidents–relationship to sleeping difficulties and occupational factors. J. Sleep Res. 11(1), 69–71 (2002)

Deep Learning Approach for COVID-19 Detection: A Diagnostic Tool Based on VGG16 and VGG19 Fardin Rahman Akash1 , Ajmiri Afrin Priniya1 , Jahani Shabnam Chadni1 , Jobaida Ahmed Shuha1 , Ismot Ara Emu1 , Ahmed Wasif Reza1(B) , and Mohammad Shamsul Arefin2,3(B) 1 Department of Computer Science and Engineering, East West University, Dhaka 1212,

Bangladesh [email protected] 2 Department of Computer Science and Engineering, Daffodil International University, Dhaka 1341, Bangladesh [email protected] 3 Department of Computer Science and Engineering, Chittagong University of Engineering and Technology, Chattogram, Bangladesh

Abstract. The coronavirus disease 2019 is a new contagious illness affecting the lungs and upper respiratory tract. It has various complications that can affect the quality of life. One of the most common factors that can be used to diagnose this illness is chest computed radiography. According to studies, deep learning can identify COVID-19 using chest radiography results. We created a CNN network to find COVID-19 in patients with Pneumonia and normal controls after a full chest X-ray. For the testing and training of the VGG16 model, we focused on its deep features. The study’s results revealed that the VGG16 model had the highest accuracy score, at 70.0021%. Keywords: VGG16 · VGG19 · Convolutional neural networks · Lung X-rays · Pneumonia

1 Introduction The outbreak of COVID-19 was first detected in Wuhan, China. Within a few months, it had already affected various parts of the world. Due to the nature of the disease and the number of people who died, it has led to a pandemic [1]. The WHO reported that over 300,000 people died globally within two months after the COVID-19 pandemic was declared. Because of the restrictions placed on people, COVID-19 severely affects every aspect of life in the world [2]. The clinical signs of an infected individual are non-specific. This means that molecular methods are required to confirm the virus [3]. COVID-19 can cause a range of illnesses from mild to severe, with high rates of morbidity and mortality. Bats are believed to be carriers of the virus, which primarily affects the respiratory system and can lead to symptoms such as fever, cough, and dyspnea. RT-PCR is commonly used © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 394–403, 2024. https://doi.org/10.1007/978-3-031-50158-6_39

Deep Learning Approach for COVID-19 Detection

395

for diagnosis but has limitations in accuracy and sensitivity [4]. Since the COVID-19 epidemic began in 2020, many papers have been released detailing remedies to the many different issues it caused [5]. Computer-based approaches have been developed to tackle various technological challenges, including the need for scientific tools to aid in the analysis of COVID-19. Early detection of the disease can save lives, and deep learning algorithms have emerged as promising tools for disease identification, reducing human error. Diagnostic methods such as chest X-rays and CT scans can also aid in detecting COVID-19. In a study involving over a thousand patients, Chinese researchers found that chest CT scans identified 97% of COVID-19 infections. COVID-19-induced pneumonia severely damages the lungs, ultimately resulting in their failure, unlike typical pneumonia which affects only a portion of the lung [6]. Our goal is to simplify the process by which patients may identify their medical issues. Because the disease can spread fast and easily from patient to patient, our key goal is for people to identify their COVID simply using their CT scan picture at home and keep others safe. This would give readers an overview of what to expect in the paper and help them understand the main points. In this case, the Introduction should include a summary of the COVID-19 outbreak and its impact, the importance of early detection and diagnosis, and the limitations of current diagnostic methods. The authors’ proposed approach using deep learning techniques, specifically comparing the VGG-16 and VGG-19 models, should also be highlighted. The potential benefits of this approach, such as its speed and practicality, should also be mentioned. Finally, the paper’s structure should also be briefly introduced, including the dataset and methods used, the results, and the authors’ conclusions and future directions.

2 Related Work Suman Chaudhary et al. [7] proposed an approach to Deep learning techniques used to identify COVID-19 in a chest X-ray image of a patient. They then trained two nets with different weights using SE-ResNext. Three techniques are used to classify the findings of X-ray images: normal, Pneumonia, and Covid. The researchers tested the proposed technique against a database of chest X-ray images for the Postgraduate Challenge. They found that it performed well with a sensitivity of 0.9592, specificity of 0.9597, and accuracy of 0.9592. Naufal Hilmizen et al. [8] proposed an approach to classify CT-Scan images into normal and COVID-19 Pneumonia categories, a combination of image recognition models and transfer learning methods, such as MobileNet, Xception, VGG16, and InceptionV3 were used. A multi-modality fusion approach combining VGG16 and ResNet50 achieved 99.87% accuracy, surpassing single-modality methods. Dimas Reynaldi et al. [9] proposed an approach to a deep learning-based approach that uses CNN architecture to identify COVID-19 infection in patients through CT-Scans. The images are preprocessed using the CLAHE method and then used to train CNN and Resnet-101 models. The approach achieved good results in accuracy and sensitivity during the testing and training phases. Ayesh Meepaganithage et al. [10] proposed an approach to develop a method that can identify patients with COVID-19 Pneumonia, non-COVID-Pneumonia, and ordinary Pneumonia using chest radiography. We then created two deep-learning models that

396

F. R. Akash et al.

could analyze the multiple images of the chest. In the COVID-19 class, the training group achieved an overall accuracy of 79%, 96% precision, and 83% recall. In the second section, the group was able to use AP view X-ray images to improve their accuracy. The suggested VGG19 model, extended from the previous model, was then used to create activation maps showing where the group could detect the virus. Amir Sorayaie Azar et al. [11] proposed an approach to identify COVID-19 patients and distinguish them from those infected with viral or bacterial Pneumonia using chest radiography images. A four-class model and a five-fold cross-validation procedure achieved the model’s accuracy. This model is lightweight and resilient; clinicians can use it to diagnose COVID-19. It has fewer parameters and training epochs and a reduced framework. Clinicians can readily recognize COVID-19 using this model as an additional diagnostic tool. Mohammad Ayyaz Azeem et al. [12] proposed an approach to analyze and evaluate the various deep-learning techniques that can be used to detect COVID-19, Pneumonia, and normal chest radiography images. Four different transfer learning methods were used for the classification tasks. The VGG16 model surpassed the competing models regarding sensitivity, specificity, and accuracy. It also performed well in detecting Pneumonia and COVID-19. Zeynep Özdemır et al. [13] proposed an approach to use neural networks and COVID19 data to detect abnormalities in images. They developed binary class classification programs based on data from NIH Chest X-ray and COVID-19 databases and used the Adaptive Sigmoid function and Contrast Limited adaptive histogram equalization to analyze the data. Three CNN models from the ImageNet library were used, and the EfficientNetB5 model performed the best. S. Bhuvana et al. [14] proposed an approach to a computer vision model trained on datasets collected from the internet was proposed to detect COVID-19 in its early stages and reduce mortality rates. The model was trained using various tools, including TensorFlow, and recommended the use of CNNs on chest X-ray scans. Zakariya A. Oraib et al. [15] proposed an approach to use machine learning and chest CT scans to identify COVID-19 patients. It used local binary patterns and the Random Forests classifier with a multiresolution approach to improve efficiency. The accuracy was evaluated with specificity and sensitivity, similar to deep learning systems. There are significant contributions in [16–20] in the field of image analysis for performing different tasks.

3 Architecture and Design of System 3.1 Dataset Description The dataset contains 11,884 chest X-ray images from children aged one to five, organized into three primary folders: Train, Test, and Validation. The dataset consists of three categories, including Pneumonia (3875), Normal (4228), and Covid (2216). The authors also mentioned that the dataset was obtained from the Kaggle website and was collected by the Guangzhou Women and Children’s Medical Center in Guangzhou. Before processing the images, the physicians reviewed the chest radiographs for quality control, and any low-quality or unreadable images were eliminated. Moreover, two

Deep Learning Approach for COVID-19 Detection

397

qualified physicians reviewed the diagnoses in the images before training the AI model. The authors also mentioned that they would manage missing values, encode categorical data, balance the dataset, and perform exploratory data analysis before choosing the most suitable predictive model with the highest accuracy score. Finally, the authors also provided information on the image preprocessing steps, which included resizing, adding multiple views, and applying various augmentation techniques. In Fig. 1, we showed input images from different perspectives.

Fig. 1. Input image from the dataset

3.2 Data Preprocessing To prepare the dataset for analysis, we will handle missing values, encode categorical data, and partition the data. We will also balance the dataset using various oversampling techniques and review it for false or redundant values. Exploratory data analysis will be conducted to gain insights from the preprocessed data. We will then use multiple deep learning models and evaluation methods like precision, recall, and f1-score to select the most suitable predictive model with the highest accuracy score. To categorize an object, we will resize the input image to 224 * 224 pixels, and the test image will be resized using the input image and uploaded again. We will also apply various image transformations such as scaling, rotation, shifting, flipping horizontally and vertically, and zooming to augment the dataset. As for citing the dataset, the authors should include the proper citation for the Kaggle website where they obtained the dataset. They should also cite the dataset source, the Guangzhou Women and Children’s Medical Center in Guangzhou. Additionally, the authors should reference any relevant papers or articles that have used the same dataset for their research. 3.3 Analysis and Findings Figure 2 shows that the target variable ratios for Covid, Normal, and Pneumonia are 35.9:40:24.2, and the bar graph shows the validation data of our dataset. The variable goal ratio of Covid, Normal, and Pneumonia is 21.5:41:37.5. The pie chart and bar chart (Fig. 3) show the train data of our dataset.

398

F. R. Akash et al.

Fig. 2. Percentage of our validation data and number of our validation data.

Fig. 3. Percentage of our train data and number of our train data.

3.4 Proposed Algorithms To train model art using our dataset, we need to be able to load it into memory. To proceed, mount Google Drive and import the RGB image dataset. Then, using our sorted dataset, we divide it into train and test sets. Furthermore, develop and build the VGG16 model. We choose it as a classification algorithm because it is one of the most common picture classification algorithms and is simple to utilize with transfer learning. Finally, train the model on a training dataset and test it on a test dataset. It should also be noted that we use VGG19 for improved performance. We have used VGG16 and added some extra layers keeping VGG16 as the base model. Here is the pseudocode of the VGG16 neural network: 1. 2. 3. 4. 5. 6. 7. 8.

9.

Import necessary libraries such as TensorFlow, Keras, etc. Define the VGG16 model in keras. Specify the input dimension and the number of classes to classify. Add the convolutional layers to the model with 64 filters of size 3 × 3 with relu activation. Add a max pooling layer of size 2 × 2 to downscale the features. Add another convolutional layer with 128 filters of size 3 × 3 with relu activation followed by another max pooling layer. Add three more convolutional layers with 256, 512, and 512 filters of size 3 × 3, all with relu activation followed by max pooling layers. Flatten the output of the last max pooling layer and pass it to a fully connected layer with 4096 neurons and relu activation, followed by another fully connected layer of 4096 neurons with relu activation. Add the final output layer with softmax activation to obtain the probability distribution of the classes.

Deep Learning Approach for COVID-19 Detection

10. 11. 12. 13. 14.

399

Compile the model with appropriate loss function, optimizer, and evaluation metric. Train the model with the training dataset and validate it with the dataset. Evaluate the model with the test dataset. Predict the classes for new data using the trained model. Save and export the trained model for future use.

3.5 Model Architecture and Design The most notable aspect of VGG16 in this regard is that rather than a large number of hyper-parameters, they focused on creating convolution layers of 3 × 3 filter with stride 1 and constantly employed the exact padding and max pool layer of 2 × 2 filter with stride 2. On the other hand, in VGG19, They also used kernels, which allowed them to cover the entire visual idea where the spatial resolution of the image was preserved using spatial padding. With stride 2, max pooling was conducted over a 2 * 2-pixel window. To execute, we used 224 * 224 pixels for both the VGG16 and VGG19 models.

4 Installation and Evaluation Outcome 4.1 Evaluation Environment We have 16 GB of RAM and 120 GB of disk space set up for us on the Google Colab laptop scenario where the envisaged system is fully constructed. The backend is Python, and for quicker development, Google’s Compute Engine has the GPU option enabled. The layers and constructed functions of the Keras Deep Learning framework are used to create the models. 4.2 Evaluation Outcome We classified a large number of Chest X-ray images to assess the performance of our model. The program’s overall effectiveness is shown in Table 1 when the chest X-ray images are divided into various groups. 4.3 Implementation We used the convolutional neural network architecture to create two unique models for the classification job. The VGG19 model and a modified version of the VGG16 model were implemented, which were previously trained to recognize distinct types of Covid19 Chest X-ray pictures. This dataset notably differs from the original dataset used to build the first VGG16 framework. It has a resolution of 224 × 224 pixels. Because we only had three classes at the time, the models’ output layer was also required to contain three layers. To finish both models’ architecture, we used convolutional, max-pooling, flattened, and thick layers. The outcomes were good. Before arriving at our final model, which had the highest accuracy, we examined a variety of architectures and attributes. The photographs in the collection were all taken from the same angle. We alter each image to present a variety of views. Versions VGG16 and VGG19 were tested.

400

F. R. Akash et al. Table 1. The models’ image detection precision for each Covid-19 chest X-ray category

Model name

Starting to learn the rate

Epoch

Batch size

Accuracy %

VGG 16

0.001

5

8

66.5268

VGG 16

0.001

16

16

70.6667

VGG 16

0.01

5

8

68.4927

VGG 16

0.01

16

16

69.2231

VGG 19

0.001

5

8

67.9923

VGG 19

0.001

16

16

70.0021

VGG 19

0.01

5

8

66.0294

VGG 19

0.01

16

16

69.0111

4.4 Performance Assessment The performance of VGG16 and VGG19 models for COVID-19 detection using chest X-rays. We have used a dataset of 22,698 images that includes COVID-19, normal, and pneumonia samples. We have also used data augmentation techniques to increase the size of the dataset. In the validation section, we report that VGG16 performs better than VGG19 with a precision of 70.6667% compared to 70.0021% (Table 2). However, both models are correct when tested on individual patients’ data. We have also provided graphs that show the precision and failure of the VGG16 model across 16 epochs. The VGG16 model takes 60 s per epoch to train, while VGG19 requires just 13 s. Our approach differs from previous studies as we have used ten angles to capture the X-ray images, resulting in a large amount of data. We also cite several other studies that have used deep learning models for COVID-19 detection using chest X-rays. Overall, having lower scores than some of the earlier research described indicates that their proposed technique provides accurate findings. Using the submitted dataset, which included COVID-19, normal, and pneumonia samples, we achieved an overall accuracy of 70.6667% using VGG16. For covid sample, we take 2358 pictures in the dataset, and for normal and Pneumonia, respectively, 15,575 and 4365 pictures. Table 2. Modeling precision and failure of VGG 19 and VGG 16 Framework

Precision in training

Failure in training

Precision in validation

Failure in validation

VGG16

0.7631

0.6253

0.6527

0.7092

VGG19

0.7253

0.6033

0.6344

0.6999

We evaluated our model in two ways: the precision of the test set and supplying an input image of a Covid-19 Chest X-ray to see if our model could detect it. The VGG19 and VGG16 models are both correct. The graphs below show the precision and failure of the VGG16 model across 16 epochs (Fig. 4).

Deep Learning Approach for COVID-19 Detection

401

Fig. 4. Model failure and precision of VGG16

When the two models are compared, the VGG16 structure needs minimal epochs to reach greater accuracy than the VGG19 structure. The VGG16 model takes 60 s per epoch to train, but the VGG19 structure requires just 13 s. As a result, the VGG19 model is substantially faster to train than the VGG16 structure. Table 3. Other works Model

Result (%)

VGG19 [9] CNN [6]

98 95.92

Robust feature Forests classifier [14] VGG16 [11]

91.3 94

From Table 3, in our paper, the best accuracy is 70.6667% in VGG16 which is 22,737 data. Our data have been transformed into extra data through augmentation. In most papers, only three angles were used to capture the photographs, and accuracy was 90–98%; however, in our paper, ten angles were used to capture the images. Using a single image with numerous dimensions resulted in an excessively enormous amount of data, which caused our paper’s accuracy to be low. Although we have low scores, our results are accurate, which is highly valuable for our medical section.

402

F. R. Akash et al.

5 Conclusion In this study, the authors compared two techniques for identifying a patient’s chest Xrays using the Deep Learning technique to identify COVID-19. By comparing VGG-16 and VGG-19, the VGG-19 produced the finest outcomes. The excellent accuracy of VGG-16 is 70.6667, and the accuracy of VGG-19 is 70.0021. This approach might serve as a significant and, using a quick procedure, patients can be identified as having COVID-19, which is speedier and more appropriate medical care. To evaluate for high computing speed, performance, and practical application of deep learning techniques, future work on a large dataset utilizing GPU will consider many more attributes. The proposed VGG-16 and VGG-19 models can circumvent these constraints. The proposed approach can potentially detect a COVID-19-positive individual in a reasonable amount of time. We anticipate that new biomarkers other than CT-Scan and X-Ray with Covid-19 pneumonia cases and larger datasets will be available in the future. Our model ought to be able to classify COVID-19 Pneumonia cases utilizing a range of various biomarkers with relevant information as the amount of biomarker modalities increases, and various biomarkers may provide supplementary information for the diagnosis of COVID-Pneumonia.

References 1. Wang, W., Xu, Y., Gao, R., Lu, R., Han, K., Wu, G., Tan, W.: Detection of SARS-CoV-2 in different types of clinical specimens. JAMA 323, 1843–1844 (2020) 2. Ji, T., Liu, Z., Wang, G., Guo, X., Lai, C., Chen, H., Huang, S., Xia, S., Chen, B., Jia, H., et al.: Detection of COVID-19: a review of the current literature and future perspectives. Biosens. Bioelectron. 166, 112455 (2020) 3. Raoult, D., Zumla, A., Locatelli, F., et al.: Coronavirus infections: epidemiological, clinical and immunological features and hypotheses. Cell Stress 4(4), 66 (2020) 4. Harmon, S.A., Sanford, T.H., Xu, S., Turkbey, E.B., Roth, H., Xu, Z., et al.: Artificial intelligence for the detection of COVID-19 pneumonia on chest CT using multinational datasets. Nat. Commun. 11(1), 4080 (2020) 5. Chu, D.K.W., Pan, Y., Cheng, S.M.S., Hui, K.P.Y., Krishnan, P., Liu, Y., et al.: Molecular diagnosis of a novel coronavirus (2019-nCoV) causing an outbreak of pneumonia. Clin. Chem. 66, 549–555 (2020) 6. Jahmunah, V., et al.: Future IoT tools for COVID-19 contact tracing and prediction: a review of the state-of-the-science. Int. J. Imaging Syst. Technol. 31(2), 455–471 (2021) 7. Chaudhary, S.: Ensemble deep learning method for Covid-19 detection via chest X-rays. http://doi.org/10.1109/EE-RDS53766.2021.9708581 8. Hilmizen, N.: The multimodal deep learning for diagnosing COVID-19 pneumonia. http:// doi.org/10.1109/ISRITI51436.2020.9315478 9. Reynaldi, D.: COVID-19 classification for chest X-ray images using deep learning and resnet101. http://doi.org/10.1109/ICOTEN52080.2021.9493431 10. Meepaganithage, A.: Detecting COVID-19 pneumonia using chest X-rays through deep learning techniques. http://doi.org/10.1109/ICARC54489.2022.9753784 11. Azar, A.S.: Lightweight method for the rapid diagnosis of coronavirus disease 2019 from chest X-ray images using deep learning technique. http://doi.org/10.1109/NSS/MIC44867. 2021.9875630

Deep Learning Approach for COVID-19 Detection

403

12. Azeem, M.A.: COVID-19 detection via image classification using deep learning on chest X-ray. http://doi.org/10.1109/EE-RDS53766.2021.9708588 13. Özdemır, Z.: Covid-19 detection in chest X-ray images with deep learning. http://doi.org/10. 1109/SIU53274.2021.9478028 14. Bhuvana, S.: Covid-19 detection using chest X-rays with image-based deep learning. http:// doi.org/10.1109/ICESC54411.2022.9885573 15. Oraib, Z.A.: Prediction of COVID-19 from chest X-ray images using multiresolution texture classification with robust local features. http://doi.org/10.1109/COMPSAC51774.2021. 00096 16. Saha, R., Debi, T., Arefin, M.S.: Developing a framework for vehicle detection, tracking and classification in traffic video surveillance. In: Vasant, P., Zelinka, I., Weber, G.W. (eds.) Intelligent Computing and Optimization. ICO 2020. Advances in Intelligent Systems and Computing, vol. 1324. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-681548_31 17. Fatema, K., Ahmed, M.R., Arefin, M.S.: Developing a system for automatic detection of books. In: Chen, J.IZ., Tavares, J.M.R.S., Iliyasu, A.M., Du, K.L. (eds.) Second International Conference on Image Processing and Capsule Networks. ICIPCN 2021. Lecture Notes in Networks and Systems, vol. 300. Springer, Cham (2022). https://doi.org/10.1007/978-3-03084760-9_27 18. Rahman, M., Laskar, M., Asif, S., Imam, O.T., Reza, A.W., Arefin, M.S.: Flower recognition using VGG16. In: Chen, J.IZ., Tavares, J.M.R.S., Shi, F. (eds.) Third International Conference on Image Processing and Capsule Networks. ICIPCN 2022. Lecture Notes in Networks and Systems, vol. 514. Springer, Cham (2022). http://doi.org/10.1007/978-3-031-12413-6_59 19. Yeasmin, S., Afrin, N., Saif, K., Imam, O.T., Reza, A.W., Arefin, M.S.: Image classification for identifying social gathering types. In: Vasant, P., Weber, GW., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). http://doi.org/10.1007/ 978-3-031-19958-5_10 20. Ahmed, F., et al.: Developing a classification CNN model to classify different types of fish. In: Vasant, P., Weber, G.W., Marmolejo-Saucedo, J.A., Munapo, E., Thomas, J.J. (eds.) Intelligent Computing & Optimization. ICO 2022. Lecture Notes in Networks and Systems, vol. 569. Springer, Cham (2023). http://doi.org/10.1007/978-3-031-19958-5_50

A Survey of Modeling the Healthcare Inventory for Emerging Infectious Diseases Tatitayakorn Limsakul(B) and Sompoap Taladgaew Department of Teacher Training in Mechanical Engineering, Faculty of Technical Education, King Mongkut’s University of Technology North Bangkok, Bangkok, Thailand [email protected], [email protected]

Abstract. Inventory management is a critical process in the healthcare industry. Challenges to the healthcare industry, such as supply shortages or overstocking, especially during the pandemic, make healthcare inventory management highly important. Efficient inventory management can help ensure quality patient care while reducing inventory costs. Over the years, several approaches and methods for modeling healthcare inventory management have been developed by researchers. This paper aims to provide an overview of inventory management (modeling) to handle inventory management problems under several constraints, such as limited budget and resources in both deterministic and probabilistic demand scenarios by focusing on the challenges posed by the COVID-19 pandemic. In this paper, techniques and methods, including Economic Order Quantity, Mathematical Optimization Models, Stochastic Programming, and Metaheuristics are presented and critically reviewed as guidance for future research. Keywords: Inventory management · Covid-19 pandemic · Optimization models · Deterministic demand · Probabilistic demand

1 Introduction In the healthcare industry, effective inventory management is critical for maintaining patient trust and providing seamless care. The drug management system is an essential component of a hospital information system, as stated by Little and Coughlan [1]. Despite its importance, inventory management in healthcare faces several challenges, such as overstocking and understocking. Overstocking can result in increased storage costs, waste, and the expiration of drugs before they are dispensed. It also exacerbates financial losses and reduces available storage space [2, 3]. Conversely, stock-outs can hinder patient care by making medical supplies and equipment unavailable when needed. The limitations of storage space and budget constraints also pose challenges for inventory management in the healthcare industry [2, 3]. Common problems in inventory management of healthcare include inaccurate demand forecasting, lack of standardization, poor inventory visibility, inefficient ordering processes, expiry of items, inadequate

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 404–413, 2024. https://doi.org/10.1007/978-3-031-50158-6_40

A Survey of Modeling the Healthcare Inventory

405

record-keeping, and inappropriate storage conditions. Addressing these issues can help healthcare organizations optimize their inventory management processes and provide high-quality care to their patients. Excessive inventory ties up capital and represents a significant financial investment that could be used elsewhere [3]. Effective inventory management helps to prevent stockouts, reduce the risk of disruptions to patient care, and ensure that medical supplies and equipment are readily available when needed. This highlights the importance of striking a balance between maintaining confidence in the system and avoiding overstocking. Inventory management in healthcare involves managing and controlling a large number and great variety of items stocked in a healthcare system [4, 5]. In the healthcare industry, striking a balance between ensuring an adequate supply of medical supplies and avoiding overstocking is a challenging task. Effective inventory management practices must be implemented to minimize the risk of stock-outs and excess inventory. This includes regularly monitoring stock levels, predicting demand, and ordering supplies in a timely manner. The use of Heuristic and Meta-Heuristic methods and regular stocktaking can also help minimize excess inventory. However, predicting demand in a healthcare system is difficult due to uncertainties and randomness, such as changes in patient conditions, dynamics in physicians’ prescriptions, and individual patient responses to treatment procedures [6, 7]. Drug shortages pose a significant challenge for healthcare institutions and often interfere with patient care. During shortages, alternate therapeutic agents are typically selected, but these agents can present challenges and raise safety concerns, leading to adverse events, medication errors, and patient complaints [8]. The timing, location, type, and amount of demand in most humanitarian settings [9, 10] bring significant challenges in developing effective inventory policies [11]. During the Covid-19 pandemic, the sudden increase in demand for medication had a significant impact on the supply chain. Medication shortages occur when the available supply of a medication is insufficient to meet current or projected demand at the patient level. The frequency of worldwide medication shortages has been rising in recent years, partly due to pharmaceutical procurement plans facing shortages of raw materials at the national level. This is due to national lockdowns in countries that produce pharmaceutical raw materials, such as the United States, China, India, and Europe [12]. Therefore, the shortage of medicines for patients is a major problem that has been widely studied [8, 13]. Different authors have developed optimization models to address the problem of medicine inventory management by considering various constraints and approaches to model uncertainty [14, 15]. The demand for healthcare items is a significant factor affecting inventory systems, and traditional forecasting techniques can predict the stationary demand [16]. However, the demand is influenced by several sources of randomness, such as patient number and treatment stage, patient condition, medication reaction, and physician recommendation [7]. In conclusion, an effective inventory management is crucial for maintaining patient satisfaction and trust in the healthcare system. By balancing the availability of supplies and the reduction of excessive inventory, healthcare facilities can establish a strong and dependable supply chain, promoting the delivery of high-quality patient care. The demand can be divided into two categories: 1) Deterministic Demand and 2) Probabilistic or Stochastic Demand.

406

T. Limsakul and S. Taladgaew

2 The Deterministic Demand: An Overview Deterministic demand in inventory management refers to demand for a product that can be predicted with certainty based on historical sales data and future demand forecasts. In contrast to probabilistic demand, which is subject to random fluctuations and external factors, deterministic demand is consistent and predictable. This type of demand is usually observed in stable markets where buying habits are consistent and well-established, such as seasonal demand for certain products or the demand for a single product in a steady-state market. In the healthcare industry, examples of deterministic demand include the demand for certain medical supplies, such as gloves and masks, during a pandemic, or the demand for seasonal. In these cases, the demand can be accurately forecasted based on historical data, and the healthcare organization can make informed decisions about how much inventory to order and when to order it. By having a good understanding of deterministic demand, healthcare organizations can avoid stock shortages, which can lead to operational disruptions and negatively impact patient care, or excessive inventory, which can result in waste and increase costs. The Economic Order Quantity (EOQ) model is a commonly used deterministic model for inventory control. This model assumes a constant and known demand for an item, constant lead time, constant unit price, inventory holding cost based on average inventory, constant ordering cost and no backorder allowed. The EOQ model has been applied in various hospital settings to improve inventory management and reduce costs, such as at Georgetown University Hospital (GUH) by Kapur and Moberg [17] in a 558 bed general hospital by Ballentine [18] at Ramathibodi Hospital by Laeiddee [19] formulas (1)  2DS (1) EOQ = H S Setup costs (per order, generally including shipping and handling) D Demand rate (quantity sold per year) H Holding costs (per year, per unit) Excess inventories in hospitals have also been studied, such as in a local hospital by Hafnika et al. [20] using continuous review policy, EOQ, reorder point, average inventory level, and ABC classification. Kritchanchai and Meesamut [21] develop the total inventory costs by EOQ model ABC classification in a large public hospital in Thailand. Other applications of deterministic demand in hospitals include blood plasma management by Ma et al. [22] using EOQ model, safety stock, and reorder point, and inventory management for a pharmaceutical company and hospital by Uthayakumar and Priyan [23] using an operations research model. Operations research has also been used to optimize a hospital’s inventory costs, such as in a hospital’s central pharmacy by Stecca et al. [24] and through the application of Model Predictive Control (MPC) by Maestre et al. [3]. These studies demonstrate the continued efforts to improve inventory management in healthcare using deterministic demand models.

A Survey of Modeling the Healthcare Inventory

407

3 The Probabilistic Demand: An Overview Probabilistic demand in inventory management takes into account the uncertainty and variability of demand for a product over time. It recognizes that demand is not always predictable with certainty and that there is a degree of randomness in the process. This randomness can be due to a variety of factors such as changes in consumer behavior, fluctuations in the economy, and unexpected events. To account for this uncertainty, organizations use probabilistic methods to model and predict demand. This involves creating a probability distribution for demand and using it to simulate different scenarios and make decisions about inventory levels. By considering the uncertainty of demand, organizations can make more informed decisions about safety stock levels, reorder points, and order quantities. This helps to ensure that inventory is always at the right level to meet customer demand, while avoiding waste and inefficiencies. The healthcare industry can benefit greatly from the application of probabilistic demand analysis in inventory management. Due to the nature of the industry, demand for medical supplies and equipment can be highly variable and unpredictable, especially in the case of emergencies or outbreaks like COVID-19. By using probabilistic methods to model demand, healthcare organizations can better prepare for fluctuations in demand and ensure that they have the necessary inventory levels to meet patient needs. For example, during a pandemic, healthcare organizations can use probabilistic demand analysis to determine the likelihood of increased demand for certain medical supplies and plan their inventory accordingly. Moreover, probabilistic demand analysis can also help healthcare organizations reduce waste and improve cost efficiency by avoiding overstocking and excessive inventory levels. By understanding the range of possible demand scenarios, organizations can make informed decisions about the optimal inventory levels to maintain and the frequency of reordering. Probabilistic demand analysis is a valuable tool for the healthcare industry in managing inventory and ensuring that critical medical supplies and equipment are available to meet patient needs. It helps organizations to be more proactive, efficient, and resilient in the face of uncertainty and variability in demand. Healthcare systems face many uncertainties, including changes in patient numbers, clinical conditions, and the availability of medicines, as highlighted by Addis et al. [25]. In response, several studies have aimed to develop models for optimizing inventory management in hospitals. 3.1 Mathematical Optimization Models Mathematical optimization models are a valuable tool for analyzing and optimizing complex systems. These models incorporate objectives or goals that are represented by mathematical functions, allowing for a systematic exploration of trade-offs and the identification of optimal solutions. In the healthcare sector, optimization models can be particularly useful for decision making and resource allocation. For example, Najafi, Ahmadi, and Zolfagharinia [26] applied optimization to blood inventory management, developing a model that considers uncertain demand and supply. Franco and AlfonsoLizarazo [27] used mixed integer programming (MIP) approaches for optimizing the pharmaceutical supply chain, taking into account elements such as demand and lead

408

T. Limsakul and S. Taladgaew

times. Additionally, Khoukhi, Bojji, and Bensouda [28] presented an inventory optimization model aimed at minimizing total cost while considering constraints such as storage space, order frequency, and service level and evaluated its performance through mathematical model and Monte Carlo simulation. Overall, the use of optimization models in healthcare can lead to improved patient care, reduced costs, and a more efficient healthcare system. Mathematical optimization models have advantages and disadvantages in decisionmaking. They offer a structured and systematic approach to decision-making and can formalize constraints and objectives, providing mathematical guarantees for optimality and feasibility. They can handle complex and large-scale problems and identify the best solution among a large number of alternatives. The other hand, the models also have several disadvantages. One significant disadvantage is that these models require a high level of mathematical expertise to formulate and solve. Furthermore, the models may not always reflect real-world complexity and uncertainty, leading to inaccuracies in the results. The computational intensive and time-consuming nature of these models, particularly for large-scale and complex problems, is another disadvantage. Finally, the solutions generated by mathematical optimization models may not always be interpretable or have practical relevance. 3.2 Stochastic Programming Stochastic Programming, also referred to as Stochastic Optimization is a mathematical framework that models decision-making under uncertainty [29]. In the area of inventory management in hospital pharmacies, it is utilized to minimize the expected total inventory costs while satisfying service level and space constraints. The complexity of the problem is due to the presence of a large number of variables, non-linearity, and stochastic constraints, where the latter refers to the requirement of maintaining a specific probability of no shortage of medicines. The utilization of two-stage stochastic programming has also been noted in the solution to the problem of allocating surgical supplies in multiple locations within a healthcare system [30, 31]. In 2014, Priyan and Uthayakumar [32] proposed a stochastic model to minimize the impact of drug shortages. Rajendran and Ravindran [33] created a mixed integer stochastic programming model to tackle demand uncertainty and presented three heuristic rules for determining the platelet ordering policy. In 2023, Meneses, Marques, and Barbosa-Póvoa [34] addressed the challenges of blood product ordering policies through a two-stage stochastic programming model that considers demand uncertainty. Stochastic Programming provides several advantages for decision-making under uncertainty. The ability to incorporate uncertainty leads to more accurate and realistic models, and the use of expected value objectives results in robust solutions. Additionally, stochastic programming can deal with multiple objectives, such as cost minimization and service level maximization, in a single model. However, stochastic programming also has some limitations. The models can be complex and computationally intensive, and the solution accuracy is dependent on the accuracy of the probability distribution used to model uncertain parameters. Additionally, the models require a significant amount of data, which can be challenging to obtain, and the validity of the model depends on the accuracy of the assumptions made about the distribution of uncertain parameters.

A Survey of Modeling the Healthcare Inventory

409

3.3 Metaheuristics In recent years, metaheuristics have become a popular tool for healthcare inventory management due to their fast and efficient solution-finding capabilities. Metaheuristics were first proposed by Glover in 1986 and have since been widely used in various optimization problems. The most commonly used metaheuristics in healthcare inventory management are the Tabu Search Algorithm (TS), Simulated Annealing Method (SAM), Ant Colony Optimization (ACO), Genetic Algorithm (GA), Emperor Penguin Optimizer (EPO), Seagull Optimizer, and Guided Local Search. The Genetic Algorithm (GA) is one of the most widely used metaheuristics in healthcare inventory management due to its ability to provide near-optimal solutions in a relatively short amount of time. This algorithm is based on Charles Darwin’s theory of natural selection and genetics, which is referred to as “Survival of the Fittest” [35]. The GA has been applied to various problems in healthcare inventory management, including reducing the total cost of the pharmaceutical supply chain from manufacturer to patient [36] and developing an optimal inventory of platelets [37]. In addition, Du, Luo, Wang, & Liu [38] proposed a model for a hospital pharmacy system using a combination of genetic algorithms and a BP neural network to improve the efficiency of pharmaceutical inventory management. The model was based on actual conditions and performed a sensitivity analysis to provide guidelines for drug inventory management in hospitals. Metaheuristics have proven to be a valuable tool in healthcare inventory management, offering a fast and efficient way to solve complex optimization problems. Despite some disadvantages, such as computational intensity and potential limitations in finding the global optimum solution, metaheuristics are a versatile and adaptable tool that can provide near-optimal solutions in a variety of inventory management scenarios. Overall, the use of metaheuristics in healthcare inventory management can lead to improved decision-making and better outcomes.

Fig. 1. Optimization model between deterministic and probabilistic demand

Figure 1, various approaches for managing healthcare inventory are presented, which includes both items with deterministic demand, such as masks and gloves, as well as items with probabilistic demand arising from uncertainty. Therefore, a joint analysis may be necessary for optimizing healthcare inventory management.

410

T. Limsakul and S. Taladgaew

4 Conclusion As the world is currently facing the Covid-19 pandemic, it continues to pose significant challenges for healthcare supply chains. Therefore, ensuring the availability of healthcare facilities by balancing demand and supply to avoid high storage costs as well as stockouts is of utmost importance. The Covid-19 situation also emphasizes that not only should patients be treated, but they must receive prompt and timely treatment. Hence, efficient healthcare inventory management will benefit patients, especially when lives are at stake, by providing quick and quality care. A refined and optimized approach to this challenge is essential to ensure that patients receive prompt and effective treatment, even during a pandemic. The techniques and methods proposed in this study can be improved for managing healthcare inventory under demand uncertainty and will be beneficial for future research in handling healthcare inventory problems, such as unpredictable demand caused by emerging infectious diseases. The conclude that to solve healthcare supply chain problems during the pandemic should consider the model of stochastic demand, which is also observed in later research focusing on stochastic demand with different problem-solving methods. In future research needs to perform experiments from equations and methods applied to real data situations to find out which method is most suitable for managing healthcare inventory. This must take into account other conditions as well, such as difficulty in actual application, calculation period, and delivery uncertainty. Table 1 appraises a number of research papers on modelling and analyzing inventory management systems in healthcare. The papers discuss types of healthcare inventory problems, existing modelling approaches, and solution methods. Table 1. Research papers of healthcare inventory Author

Ballentine et al. [18] Kapur et al. [17]

Published Demand Methodology year Deterministic Stochastic EOQ Mathematical Stochastic Genetic model programming algorithm √ √ 1976 1987

Laeiddee [19] 2010 Kelle et al. [30]

√ √

Uthayakumar 2013 et al. [23] Priyan et al. [32]

2014

Kritchanchai et al. [21]

2015

Hafnika et al. 2016 [20]

√ √

2012

Ma et al. [22] 2013

√

√

√ √

√

√ √

√

√

√

√

√

(continued)

A Survey of Modeling the Healthcare Inventory

411

Table 1. (continued) Author

Stecca et al. [24]

Published Demand Methodology year Deterministic Stochastic EOQ Mathematical Stochastic Genetic model programming algorithm √ √ 2016

Najafi et al. [26]

2017

Rajendran et al. [33]

2017

Maestre et al. 2018 [3] Khoukhi et al. [28]

2019

Rajendran et al. [37]

2019

Franco et al. [27]

2020

Du et al. [38]

2020

Nasrollahi et al. [36]

2021

Meneses et al. [34]

2023

√

√

√ √

√ √

√

√

√ √

√ √

√

√

√

√

√

√

References 1. Little, J., Coughlan, B.: Optimal inventory policy within hospital space constraints. Health Care Manag. Sci. 11(2), 177–183 (2008) 2. Bijvank, M., Vis, I.F.: Inventory control for point-of-use locations in hospitals. J. Oper. Res. Soc. 63, 497–510 (2012) 3. Maestre, J.M., Fernández, M.I., Jurado, I.J.C.E.P.: An application of economic model predictive control to inventory management in hospitals. Control Eng. Pract. 71, 120–128 (2018) 4. Gebicki, M., Mooney, E., Chen, S.J., Mazur, L.M.: Evaluation of hospital medication inventory policies. Health Care Manag. Sci. 17, 215–229 (2014) 5. Nicholson, L., Vakharia, A.J., Erenguc, S.S.: Outsourcing inventory management decisions in healthcare: models and application. Eur. J. Oper. Res. 154(1), 271–290 (2014) 6. Montoya, R., Netzer, O., Jedidi, K.: Dynamic allocation of pharmaceutical detailing and sampling for long-term profitability. Mark. Sci. 29(5), 909–924 (2010) 7. Vila-Parrish, A.R., Ivy, J.S., King, R.E., Abel, S.R.: Patient-based pharmaceutical inventory management: a two-stage inventory and production model for perishable products with Markovian demand. Health Syst. 1(1), 69–83 (2012) 8. McLaughlin, M., et al.: Effects on patient care caused by drug shortages: a survey. J. Manag. Care Pharm. 19(9), 783–788 (2013) 9. Hughes, D., McGuire, A.: Stochastic demand, production responses and hospital costs. J. Health Econ. 22(6), 999–1010 (2003) 10. Darmian, S.M., Fattahi, M., Keyvanshokooh, E.: An optimization-based approach for the healthcare districting under uncertainty. Comput. Oper. Res. 135, 105425 (2021)

412

T. Limsakul and S. Taladgaew

11. Whybark, D.C.: Issues in managing disaster relief inventories. Int. J. Prod. Econ. 108(1–2), 228–235 (2007) 12. Goodarzian, F., Taleizadeh, A.A., Ghasemi, P., Abraham, A.: An integrated sustainable medical supply chain network during COVID-19. Eng. Appl. Artif. Intell. 100, 104188 (2021) 13. Caulder, C.R., Mehta, B., Bookstaver, P.B., Sims, L.D., Stevenson, B., South Carolina Society of Health-System Pharmacists: Impact of drug shortages on health system pharmacies in the Southeastern United States. Hosp. Pharm. 50(4), 279–286 (2015) 14. Fox, E.R., Sweet, B.V., Jensen, V.: Drug shortages: a complex health care crisis. In: Mayo Clinic Proceedings, vol. 89, no. 3, pp. 361–373. Elsevier, Amsterdam (2014) 15. Kwon, I.W.G., Kim, S.H., Martin, D.G.: Healthcare supply chain management; strategic areas for quality and financial improvement. Technol. Forecast. Soc. Chang. 113, 422–428 (2016) 16. Ramirez, A.L., Jurado, I., Garcia, M.F., Tejera, B.I., Llergo, J.D.P., Torreblanca, J.M.: Optimization of the demand estimation in hospital pharmacy. In: Proceedings of the 2014 IEEE emerging technology and factory automation (ETFA), pp. 1–6. IEEE (2014) 17. Kapur, R., Moberg, C.: Evaluating inventory turns for a hospital environment. Comput. Ind. Eng. 13(1–4), 73–77 (1987) 18. Ballentine, R., Ravin, R.L., Gilbert, J.R.: ABC inventory analysis and economic order quantity concept in hospital pharmacy purchasing. Am. J. Hosp. Pharm. 33(6), 552–555 (1976) 19. Laeiddee, C.: Improvement of re-order point for drug inventory management at Ramathibodi Hospital. Doctoral dissertation, Mahidol University (2010) 20. Hafnika, F., Farmaciawaty, D.A., Adhiutama, A., Basri, M.H.: Improvement of inventory control using continuous review policy in a local hospital at Bandung City, Indonesia. Asian J. Technol. Manag. 9(2), 109 (2016) 21. Kritchanchai, D., Meesamut, W.: Developing inventory management in hospital. Int. J. Supply Chain Manag. 4(2), 11–19 (2015) 22. Ma, J., Lei, T., Okudan, G.E.: EOQ-based inventory control policies for perishable items: the case of blood plasma inventory management. In: IIE Annual Conference, Proceedings, pp. 1683–1692. Institute of Industrial and Systems Engineers (IISE) (2013) 23. Uthayakumar, R., Priyan, S.: Pharmaceutical supply chain and inventory management strategies: optimization for a pharmaceutical company and a hospital. Oper. Res. Health Care 2(3), 52–64 (2013) 24. Stecca, G., Baffo, I., Kaihara, T.: Design and operation of strategic inventory control system for drug delivery in healthcare industry. IFAC-PapersOnLine 49(12), 904–909 (2016) 25. Addis, B., Carello, G., Grosso, A., Lanzarone, E., Mattia, S., Tànfani, E.: Handling uncertainty in health care management using the cardinality-constrained approach: advantages and remarks. Oper. Res. Health Care 4, 1–4 (2015) 26. Najafi, M., Ahmadi, A., Zolfagharinia, H.: Blood inventory management in hospitals: considering supply and demand uncertainty and blood transshipment possibility. Oper. Res. Health Care 15, 43–56 (2017) 27. Franco, C., Alfonso-Lizarazo, E.: Optimization under uncertainty of the pharmaceutical supply chain in hospitals. Comput. Chem. Eng. 135, 106689 (2020) 28. Khoukhi, S., Bojji, C., Bensouda, Y.: Optimal inventory control policy for a hospital case. In: 2019 5th International Conference on Optimization and Applications (ICOA), pp. 1–5. IEEE (2019) 29. Birge, J.R., Louveaux, F.: Introduction to Stochastic Programming. Springer Science & Business Media, Berlin (2011) 30. Kelle, P., Woosley, J., Schneider, H.: Pharmaceutical supply chain specifics and inventory solutions for a hospital case. Oper. Res. Health Care 1(2–3), 54–63 (2012) 31. Ahmadi, E., Masel, D.T., Hostetler, S.: A robust stochastic decision-making model for inventory allocation of surgical supplies to reduce logistics costs in hospitals: a case study. Oper. Res. Health Care 20, 33–44 (2019)

A Survey of Modeling the Healthcare Inventory

413

32. Priyan, S., Uthayakumar, R.: Optimal inventory management strategies for pharmaceutical company and hospital supply chain in a fuzzy–stochastic environment. Oper. Res. Health Care 3(4), 177–190 (2014) 33. Rajendran, S., Ravindran, A.R.: Platelet ordering policies at hospitals using stochastic integer programming model and heuristic approaches to reduce wastage. Comput. Ind. Eng. 110, 151–164 (2017) 34. Meneses, M., Marques, I., Barbosa-Póvoa, A.: Blood inventory management: ordering policies for hospital blood banks under uncertainty. Int. Trans. Oper. Res. 30(1), 273–301 (2023) 35. Holland, J.H.: Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence. MIT Press (1992) 36. Nasrollahi, M., Razmi, J.: A mathematical model for designing an integrated pharmaceutical supply chain with maximum expected coverage under uncertainty. Oper. Res. Int. J. 21(1), 525–552 (2019). https://doi.org/10.1007/s12351-019-00459-3 37. Rajendran, S., Ravindran, A.R.: Inventory management of platelets along blood supply chain to minimize wastage and shortage. Comput. Ind. Eng. 130, 714–730 (2019) 38. Du, M., Luo, J., Wang, S., Liu, S.: Genetic algorithm combined with BP neural network in hospital drug inventory management system. Neural Comput. Appl. 32(7), 1981–1994 (2020)

MEREC-MABAC Based-Parametric Optimization of Chemical Vapour Deposition Process for Diamond-Like Carbon Coatings Sellamuthu Prabhukumar1 , Jasgurpeet Singh Chohan2 , and Kanak Kalita3(B) 1 Department of Mechanical Engineering, Presidency University, Bangalore, India

[email protected]

2 Department of Mechanical Engineering, University Centre for Research and Development,

Chandigarh University, Mohali, India 3 Department of Mechanical Engineering, Vel Tech Rangarajan Dr, Sagunthala R&D Institute of

Science and Technology, Avadi, India [email protected]

Abstract. This research paper presents an application of the Multi-attributive Border Approximation Area Comparison (MABAC) method combined with the Method based on the Removal Effects of Criteria (MEREC) for the parametric optimization of the Chemical Vapor Deposition (CVD) process in Diamond-like Carbon (DLC) coatings. A decision matrix is formulated using a case study from the literature. Four response parameters namely, Hardness (H), Young’s modulus (E), Coefficient of Friction (COF), and Wear Rate (WR) are considered. The weight allocation for these response parameters is calculated using five different methods, namely MEREC, mean weight (Mean), Standard deviation (StDev), Entropy, and Criteria Importance Through Intercriteria Correlation (CRITIC) method. The MABAC method was employed to obtain the optimal parametric combination for the DLC coatings. Results showed a clear superior combination of the CVD process parameters can be achieved using the MEREC-MABAC methodology. Thus, the study successfully demonstrates the effectiveness of the MEREC-MABACbased approach for the simultaneous optimization of multiple responses in the CVD process for DLC coatings. Keywords: Diamond-like carbon coatings · Chemical vapor deposition · Parametric optimization · MABAC · MEREC · Multi-criteria decision making

1 Introduction Diamond-like Carbon (DLC) coatings have been widely researched due to their exceptional properties, such as high hardness, low wear rate, and low coefficient of friction, which make them suitable for various applications in industries such as automotive, aerospace, and biomedical. The Chemical Vapor Deposition (CVD) process is commonly used to develop these coatings, but the quality of the coatings is significantly influenced by the process parameters. Among the numerous CVD parameter, the H2 © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 414–422, 2024. https://doi.org/10.1007/978-3-031-50158-6_41

MEREC-MABAC Based-Parametric Optimization of Chemical

415

flow rate, C2 H2 flow rate and deposition temperature (T ) are found to be some of the most significant ones. Therefore, optimizing these parameters is crucial to achieving coatings with desired properties. In recent years, various research studies have been conducted to investigate the optimization of deposition parameters for DLC coatings using different techniques and methodologies. Ghadai et al. [1] employed a thermal CVD process and used input parameters such as temperature (T ), H2 flow rate and C2 H2 flow rate to optimize hardness (H) and Young’s modulus (E). Jatti et al. [2] investigated the optimization of H , E, and ID/IG ratio using the Inductively Coupled PECVD process. They considered input parameters like voltage (V ), frequency (f ), pressure (P), and gas composition. An L9 experimental design was used with nine experiments, and the Taguchi methodology was applied for optimization. Ghadai et al. [3] used the CVD process and input parameters V , f , P, and gas composition to optimize H , E, and ID/IG ratio. They employed an L9 experimental design with nine experiments and utilized the Grey fuzzy logic. Singh and Jatti [4] focused on the IC-PECVD process, optimizing H and E with input parameters like V , f , P and gas composition. They used an L9 experimental design with nine experiments and the Taguchi methodology for optimization. Ghadai et al. [5] used the PECVD process to optimize H with input parameters like T, H2 flow rate, and C2 H2 flow rate. They employed a CCD experimental plan with 20 experiments and utilized a single-objective GA for optimization. Ebrahimi et al. [6] employed a CVD process with input parameters T and H2 flow rate to optimize the wear rate (WR) and the coefficient of friction (COF). They used a CCD experimental plan with 13 experiments and applied the desirability function approach for optimization. Ebrahimi et al. [7] utilized a CVD process and input parameters T , duty cycle, H2 flow rate, and argon/methane flow ratio to optimize WR, wear durability, and H. They employed a CCD experimental plan with 23 experiments and used the desirability function approach for optimization. Kumar and Swain [8] used a thermal CVD process with input parameters T, H2 flow rate, and N2 flow rate to optimize H , E, and ID/IG ratio. Pancielejko et al. [9] used a modified cathodic vacuum arc method with input parameters V, argon pressure, coating thickness (t), and thickness of chromium interlayer (tcr ) to optimize H and WR. They employed an L9 experimental design with nine experiments and utilized the Taguchi methodology for optimization. Czyzniewski et al. [10] investigated the optimization of H , WR, adhesion, and H /E parameter using sputtering. They used input parameters like V , C2 H2 flow rate, t, and tcr . An L9 experimental design was employed with nine experiments, and the Taguchi methodology was applied for optimization. From the literature review, it is found that several optimization methods have been applied in the literature for multi-objective problems. However, there is no application of newer methods like MEREC and MABAC in CVD process optimization. Thus, in this study, the Multi-attributive Border Approximation Area Comparison (MABAC) method, combined with the Method based on the Removal Effects of Criteria (MEREC) is employed for parametric optimization of the CVD process for DLC coatings. The main aim is to find the optimal CVD process parameters that yield the best compromise between Hardness (H ), Young’s modulus (E), Coefficient of Friction (COF), and Wear Rate (WR).

416

S. Prabhukumar et al.

2 Methodology 2.1 MEREC MEREC is a method for determining the weights of various criteria in multi-criteria decision-making (MCDM) problems [11]. MEREC focuses on the removal effect of each criterion on the alternative’s performance. Criteria with higher effects on the performances receive greater weights. A logarithmic measure calculates alternatives’ performances, and the absolute deviation measure identifies the effects of removing each criterion. The pseudo-code for MEREC is as follows: 1. Define the decision matrix (X ) with elements xij . 2. Normalize the decision matrix (N ) using nxij ⎧ min xkj ⎨ k xij if j ∈ B nxij = xij ⎩ max x if j ∈ C k

(1)

kj

where B is the set of beneficial criteria and C is the set of non-beneficial (cost) criteria. 3. Calculate overall performance Si for each alternative i using a logarithmic measure ⎞⎞ ⎛ ⎛   1 (2) Si = ln⎝1 + ⎝ ln nxij ⎠⎠ m j

4. Calculate performance Sij  of each alternative i by removing criterion j ⎞⎞ ⎛ ⎛    1 ln nx ⎠⎠ Sij = ln⎝1 + ⎝ ik m

(3)

k,k=j

5. Compute the summation of absolute deviations Ej for each criterion j  Ej = |Sij − Si |

(4)

i

6. Determine the final weights wj of the criteria Ej wj =  k Ek

(5)

2.2 MABAC The MABAC method is an MCDM technique designed to evaluate, rank, and select the best alternatives among a set of decision alternatives based on multiple criteria [12]. MABAC is particularly effective in dealing with complex decision-making problems that involve conflicting criteria, as it incorporates the concept of border approximation area to determine the relative importance of each alternative. This approach facilitates the ranking of alternatives by comparing their proximity to an ideal solution, which is represented by a border approximation area. The pseudo-code for MABAC is as follows:

MEREC-MABAC Based-Parametric Optimization of Chemical

417

1. Develop the decision matrix (X ) with m alternatives and n criteria. xij are the elements of X . 2. Normalize X to form the normalized matrix R (with elements rij ) using the following rules rij =

xij − xj− xj+ − xj−

rij =

for benefit criteria

xij − xj+ xj− − xj+

for cost criteria

xj+ and xj− are the maximum and minimum values of the J th criterion. 3. Compute the weighted normalized decision matrix V (with elements vij )

 vij = wj · rij + 1

(6)

(7)

(8)

wj is the weight of the J th criterion. 4. Compute the border approximation area (BAA) matrix B (with elements bj )  bj =

m 

1/ m vij

(9)

i=1

5. Compute the distance matrix of alternatives (Q) from the BAA. qij are the elements of Q. Q =V −B

(10)

6. Compute the criteria function (Si ) values and ranking the alternatives: Si =

n 

qij , j = 1, 2, . . . , n, i = 1, 2, . . . , m

(11)

j=1

7. Rank the alternatives in descending order of Si values.

3 Problem Description In this work, the objective is to select the optimal CVD process parameters to develop an optimized DLC coating. The challenge is to find a suitable compromise solution wherein multiple responses are looked upon and optimized simultaneously. In the context of this study, four response parameters, namely Hardness (H ), Young’s modulus (E), Coefficient of Friction (COF) and Wear Rate (WR) need to be optimised simultaneously. The CVD deposition process parameters are H2 flow rate, C2 H2 flow rate and deposition temperature (T ). Based on a central composite design of experiments, 15 experiments were conducted by Kalita et al. [13]. Those experiments are used as the decision matrix

418

S. Prabhukumar et al.

in this study for further analysis. Thus, in the context of this study, the decision matrix (D) iexpressed as, ⎤ ⎡ H E COF WR ⎥ A1 ⎢ ⎢ 13.37 141.74 0.24 0.00084 ⎥ ⎢ A2 ⎢ 20.56 272.73 0.14 0.00033 ⎥ ⎥ ⎥ A3 ⎢ ⎢ 14.31 146.36 0.21 0.00072 ⎥ ⎥ ⎢ A4 ⎢ 23.22 289.65 0.074 0.00035 ⎥ ⎥ ⎢ A5 ⎢ 16.69 170.52 0.146 0.00065 ⎥ ⎥ ⎢ ⎥ A6 ⎢ ⎢ 39.35 350.24 0.06 0.00012 ⎥ ⎥ 18.21 183.73 0.185 0.00056 A7 ⎢ ⎥ ⎢ (8) D= ⎥ 20.11 250.36 0.159 0.00045 A8 ⎢ ⎥ ⎢ ⎥ A9 ⎢ ⎢ 24.59 292.35 0.086 0.00031 ⎥ ⎥ ⎢ A10 ⎢ 22.48 283.75 0.105 0.000268 ⎥ ⎥ ⎢ A11 ⎢ 34.61 312.18 0.074 0.000132 ⎥ ⎥ ⎢ ⎥ A12 ⎢ ⎢ 21.05 275.48 0.16 0.00038 ⎥ ⎢ A13 ⎢ 36.33 325.49 0.094 0.000128 ⎥ ⎥ A14 ⎣ 23.22 287.77 0.142 0.00032 ⎦ A15 30.12 298.56 0.125 0.00025 As indicated earlier, in this paper, the MABAC method is used for multi-criteria decision-making. The weights for the four criteria are calculated using the MEREC method. However, for the sake of comprehensive comparison, the analysis is also carried out using other weight allocation methods namely, mean weight (Mean), Standard deviation (StDev), Entropy and Criteria Importance Through Intercriteria Correlation (CRITIC) method.

4 Results and Discussion 4.1 Multi-criteria Decision Making Initially, the weights for the four criteria i.e., Hardness (H ), Young’s modulus (E), Coefficient of Friction (COF) and Wear Rate (WR) are calculated by using the five different weight allocation methods. Figure 1 shows the weights allocated by the various weight allocation methods. It is observed that StDev has almost the same allocation as the mean method. However, the Entropy method is seen to have allocated skewed weights with excessive weightage to WR response. The MABAC calculations are carried out using all these weights and the Q-values are derived. The correlation between the solutions by the various weighted MABACs is shown in Fig. 2, which shows that there is a 100% correlation among the methods for this case study. This indicates that the parametric combination in the CCD-based CVD experiments is such that there is a clear superior combination. Figure 3 shows the changes in the Q-values with respect to the various parametric combination in the CCD-based CVD experimental dataset. It should be noted here that the Q-value can be thought of as a ‘combined proxy index’ for the goal of simultaneous

MEREC-MABAC Based-Parametric Optimization of Chemical

419

Fig. 1. Weights assigned to different criteria as per various weight allocation methods

Fig. 2. Correlation among the various weight allocation methods

maximization of H and E while minimizing COF and WR. Thus, the higher the Qvalue, the better the compromise solution. It is observed that irrespective of the weight allocation method, the Q-values follow a similar trend. Experiment number 6 is seen to be a clear winner that represents a good parametric combination. The various parameter value for this experiment is 60 sccm of H2 flow rate, 2.5 sccm of C2 H2 flow rate and deposition temperature (T ) of 800 ˚C. 4.2 Parametric Optimization of CVD The MABAC Q-values are aggregated level-wise for each of the three process parameters to find out the optimal parametric combination. A higher value of aggregated Q-value corresponds to a better parametric combination. Figure 4 shows the influence of the H2 the flow rate on Q-values. It is observed that as the H2 flow rate is increased the Q-values improve. However, at 80 sccm of H2 flow rate and 95 sccm of H2 flow rate, the Q-values are similar.

420

S. Prabhukumar et al.

Fig. 3. Variation of MABAC Q-values with respect to CCD-based CVD experiments

Fig. 4. Effect of H2 flow rate on aggregated Q-values

Figure 5 shows the influence of the C2 H2 flow rate on the MABAC Q-values. The Q-values monotonically decrease as the C2 H2 flow rate increases. 2.5 sccm of C2 H2 flow rate is found to be the most optimal. The drop in Q-value between 2.5 sccm of C2 H2 flow rate and 9.5 sccm of C2 H2 flow rate is 151.92%. This indicates the importance of choosing the optimal parameters for achieving the best performance from the DLC coatings. Figure 5 shows the influence of the deposition temperature (T ) on Q-values. A higher deposition temperature (T ) is seen to be beneficial for achieving a better optimized DLC coating. Thus, as per the MEREC-MABAC analysis, a deposition temperature (T ) of 900 ˚C is most beneficial in achieving the optimized DLC (Fig. 6).

MEREC-MABAC Based-Parametric Optimization of Chemical

421

Fig. 5. Effect of C2 H2 flow rate on aggregated Q-values

Fig. 6. Effect of deposition temperature (T ) on aggregated Q-values

5 Conclusions In this research, MEREC-MABAC-based approach was applied for the parametric optimization of the CVD process for DLC coatings. The optimal parametric combination for achieving the best compromise between Hardness (H), Young’s modulus (E), Coefficient of Friction (COF) and Wear Rate (WR) was determined. The study demonstrated that irrespective of the weight allocation method used, the parametric combination in the CCD-based CVD experiments showed a clear superior combination, with experiment number 6 having the highest Q-value. The analysis revealed that higher H2 flow rate, lower C2 H2 flow rate and higher deposition temperature (T ) were beneficial for achieving the optimized DLC coatings. The results of this study can provide useful insights for researchers and practitioners

422

S. Prabhukumar et al.

in the field of DLC coatings. The proposed MEREC-MABAC-based approach can be extended to other multi-objective optimization problems in various fields.

References 1. Ghadai, R.K., Kalita, K., Mondal, S.C., Swain, B.P.: Genetically optimized diamond-like carbon thin film coatings. Mater. Manuf. Process 34, 1476–1487 (2019) 2. Jatti, V.S., Laad, M., Singh, T.P.: Taguchi approach for diamond-like carbon film processing. Proc. Mater. Sci. 6, 1017–1023 (2014) 3. Ghadai, R.K., Das, P.P., Shivakoti, I., Mondal, S.C., Swain, B.P.: Grey fuzzy logic approach for the optimization of DLC thin film coating process parameters Using PACVD technique. IOP Conf. Ser. Mater. Sci. Eng. 1–6 (2017) 4. Singh, T.P., Jatti, V.S.: Optimization of the deposition parameters of DLC coatings with the IC-PECVD method. Part Sci. Technol. 33(2), 119–123 (2015) 5. Ghadai, R.K., Kalita, K., Mondal, S.C., Swain, B.P.: PECVD process parameter optimization: towards increased hardness of diamond-like carbon thin films. Mater. Manuf. Process. 33, 1905–1913 (2018) 6. Ebrahimi, M., Mahboubi, F., Naimi-Jamal, M.R.: RSM base study of the effect of deposition temperature and hydrogen flow on the wear behavior of DLC films. Tribol. Int. 91, 23–31 (2015) 7. Ebrahimi, M., Mahboubi, F., Naimi-Jamal, M.R.: Optimization of pulsed DC PACVD parameters: toward reducing wear rate of the DLC films. Appl. Surf. Sci 389, 521–531 (2016) 8. Kumar, D., Swain, B.: Investigation of structural and mechanical properties of silicon carbonitride thin films. J. Alloys Compd. 789, 295–302 (2019) 9. Pancielejko, M., Czy˙zniewski, A., Zavaleyev, V., Pander, A., Wojtalik, K.: Optimization of the deposition parameters of DLC coatings with the MCVA method. Arch. Mater. Sci. Eng. 54(2), 60–67 (2012) 10. Czyzniewski, A.: Optimising deposition parameters of W-DLC coatings for tool materials of high-speed steel and cemented carbide. Vacuum 86(12), 2140–2147 (2012) 11. Kalita, K., Ghadai, R.K., Chakraborty, S.: Parametric optimization of CVD process for DLC thin film coatings: a comparative analysis. S¯adhan¯a 47(2), 57 (2022) 12. Keshavarz-Ghorabaee, M., Amiri, M., Zavadskas, E.K., Turskis, Z., Antucheviciene, J.: Determination of objective weights using a new method based on the removal effects of criteria (MEREC). Symmetry 13(4), 525 (2021) ´ 13. Pamuˇcar, D., Cirovi´ c, G.: The selection of transport and handling resources in logistics centers using multi-attributive border approximation area comparison (MABAC). Expert Syst. Appl. 42(6), 3016–3028 (2015)

Author Index

A Addawe, Rizavel C. 125 Adhikari, Sujita 196 Ahmed, Ikbal 334 Akash, Fardin Rahman 394 Akash, Rashik Shahriar 181 Al Fahim, Hafiz 181 Al Munem, Abdullah 288 Alanas, Rostum Paolo B. 125 Alexander, 244 Ali, Mohammed Nadir Bin 379 Anisimov Alexander, A. 153 Ansari, Sarfarazali 95 Anwar, Jawed 95 Arefin, Mohammad Shamsul 181, 262, 288, 359, 379, 394 Arein, Mohammad Shamsul 368 Aruntippaitoon, Worapat 253 B B.S., Goryachkin 76 Bandyopadhyay, Tarun Kanti 170 Banik, Anirban 145, 170 Banik, Nayan 334 Bekda¸s, Gebrail 350 Bhuiyan, Touhid 181, 379 Biswal, Sushant Kumar 170 Boojhawon, Ravindra 65 Buleeva, S. 217

E Emu, Ismot Ara

134

65

394

G Garhwal, Anil 145, 170 Garhwal, Sunil 145, 170 Gukhool, Oomesh 65 H Haque, Rafid Mahmud 368 Higashi, Masaki 84 Hoque, Md Mahmudul 334 Hoque, Mohammed Moshiul 334 I Ignatkin, I. Yu. 153 Imam, Omar Tawhid 288 Injamul Haque, Md. 379 Ishkin, P. 163 Islam, Efte Kharul 262 Islam, Saiful 379 J Jabed Hosen, Md. 379 Jarin, Tanin Mohammad Jeong, Seung Ryul 33

C Chadni, Jahani Shabnam 394 Chanta, Sunarin 14, 253 Chhetri, Shanti Devi 3, 22 Chohan, Jasgurpeet Singh 104, 414 Choppradit, Pakcheera 314 D Daus, Yu. 163 Doctolero, Angela Ronice A. 125 Domínguez-Miranda, Sergio Arturo

Doomah, Mohammad Zaheer

181

K Kalita, Kanak 104, 414 Kazantsev Sergey, P. 153 Khan, Raashid 95 Kildeev, T. A. 207 Kildeev, T. 153 Kombarov, Volodymyr 277 Kononenko, A. S. 207 Kumar, Devesh 3 Kumar, Santosh 301

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 P. Vasant et al. (Eds.): ICO 2023, LNNS 855, pp. 423–425, 2024. https://doi.org/10.1007/978-3-031-50158-6

424

L Libatique, Criselda P. 125 Limsakul, Tatitayakorn 404 Litvinchev, Igor 277 M Maha, Lamyea Tasneem 368 Manshahia, Mukhdeep Singh 236 Mashkov, S. 163 Maza, Guinness G. 125 Meem, Afsana Nur 262 Mim, Faria Tabassum 359 Moin, Nafisha Binte 288 Moskvin, Valery 115 Munapo, Elias 301 Muradyan, S. 217 N Nakata, Kazuhide 84 Nigdeli, Sinan Melih 350 Nikolsky, D. R. 76 Nikulina, E. 217 Nodi, Lamya Ishrat 262 Nyamugure, Philimon 301 O Octavia, Tanti 223 P Pagunsan, Clarence Kyle L. 125 Palit, Herry Christian 244 Panchenko, Vladimir 170 Pankratov, Oleksandr 277 Pathan, Azazkhan Ibrahimkhan 95 Phimsiri, Sasin 314 Pishchaeva, K. 217 Plankovskyy, Sergiy 277 Poudel, Sudip 33 Prabhukumar, Sellamuthu 104, 414 Priniya, Ajmiri Afrin 394 R Rahman, Oshin Nusrat 368 Rajkarnikar, Neesha 33 Ranabhat, Deepesh 3, 22, 45, 196 Randhawa, Jasleen 236 Rebolledo, Jonathan 56 Reza, Ahmed Wasif 181, 262, 288, 359, 368, 379, 394

Author Index

Rodriguez-Aguilar, Roman Romanova, Tetyana 277

56, 134

S Safa, Noor Fabi Shah 368 Saha, Arpita 359 said, Saif 95 Saiful, Md. 359 Sangsawang, Ornurai 14 Santoso, Leo Willyanto 342 Sapkota, Pradeep 3, 22, 45 Saproshina, A. 217 Sergeeva, N. A. 207 Serov, A. V. 153 Serov, N. V. 153 Sharma, Sandeep Kumar 145 Sharma, Shruti 145 Shestov, Dmitry 115 Shilin, Denis 115 Shrestha, Deepanjal 33 Shrestha, Deepmala 33 Shuha, Jobaida Ahmed 394 Sidek, Lariyah Mohd 95 Singh, Aditya 324 Skorokhodov, D. M. 153 Sookpong, Satida 314 Soosor, Nooswaibah Binti Nooroodeen Sripathomswat, Kanokporn 253 Sukhobokov, A. A. 76 Sultana, Shamima 288 Suttichaya, Vasin 314 Syrkin, V. 163 T Taladgaew, Sompoap 404 Tapu, Md. Abu Bakar Siddiq 181 Tasnim, Nafisa 359 Tawanda, Trust 301 Thamwiwatthana, Ek 314 Tipchareon, Nattawat 253 Tomal, Minhazul Amin 262 Tosawadi, Teepakorn 314 Trirattanasarana, Itiphong 253 Tsegelnyk, Yevgen 277 Tuhin, Rashedul Amin 368 Tunga, Tiovitus Flomando 223 U Ukidve, Seema

236

65

Author Index

Utintu, Chaitat 314 Utsumi, Yoshimasa 84 V Vashisth, Kamal Kant 45 Vasilev, S. 163 Vasiliev, Alexey 115 Verma, Narinder 22, 196

425

Y Yadav, Ramsagar 236 Yu, Ignatkin I. 207 Yücel, Melda 350

Z Zante, Kendrick Jules G.

125