Trends in Sustainable Smart Cities and Territories 9783031369575, 3031369572

This book presents the latest scientific and technical advances in the fields of Smart Cities and Smart Territories. It

139 1 34MB

English Pages 506 Year 2023

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Preface
Organization
Contents
Main Track
Motorized Mobility on a Latin American University Campus: A Preliminary Study Focused on Sustainability
1 Introduction
2 Methodology
3 Results and Discussions
3.1 Global Characterization of University Campus Mobility
3.2 Characterization of University Campus Motorized Vehicle Fleet
4 Conclusions
References
Smart Urban Governance Through Geoinformation: The Importance of Geoportals for City Interoperability
1 Introduction
2 The Concept of Smart Urban Governance and the Approach of Law
3 The Use of Geoinformation by Cities Through Geoportals
4 Interoperability as a Principle for Smart Urban Governance
5 Conclusions
References
Digital Twin Application Methodology for the Improvement of Production and Service Systems. Application to Waste Management Processes
1 Introduction
2 Methodology
2.1 Digital Twins, Definitions and Characteristics
2.2 Waste Management
2.3 Methodological Development
3 Results
3.1 Digital Twin Definition
3.2 Architecture and Framework
3.3 Methodological Proposal
4 Future Work
5 Conclusion
References
Adaptive and Intelligent Edge Computing Based Building Energy Management System
1 Introduction and Motivation
2 State of Art
3 Architecture
3.1 Layer architecture
3.2 Multi-agent Architecture
3.3 Analysis Architecture
4 Case of Use
4.1 Implementation Architecture
5 Conclusion and Future Work
References
A Novel Aging-Based Proof of Stake Consensus Mechanism
1 Introduction
2 Related Works
3 Methodology
3.1 Problem Statement
3.2 Proposed Method
4 Simulation and Result Analysis
4.1 Evaluation Parameters
4.2 Fairness Analysis of Aging-Based PoS
4.3 Decentralization Analysis of Aging-Based PoS
4.4 Security Analysis of Aging-Based PoS
5 Conclusion
References
Manizales, Smart and Sustainable Tourist Destination
1 Introduction
2 Methodology
3 Revised Standards for Measuring Smart and Sustainable Tourist Cities
4 Manizales, Context and Vision of the City
5 Model for Manizales as a Smart Sustainable Tourist Destination
6 Action Plan
7 Conclusion and Future Work
References
Development of an IoT Network for Urban Orchards in High Vulnerability Areas in Colombia
1 Introduction
2 Context
3 Implementation of IoT Network
4 Results and Contribution
4.1 Multiple Linear Regression Models
4.2 Genetic Algorithms
5 Conclusions
References
DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture
1 Introduction
2 Related Work
3 Materials and Method
3.1 Data Set for Training
3.2 Processing of the Dataset
3.3 Validation with Real-Time Traffic
3.4 Cloud Architecture
4 Results
5 Conclusions
6 Limitations and Future Work
References
Smart Cities Using Crowdsensing and Geoferenced Notifications
1 Introduction
2 Methodology
3 Results
4 Discussion
4.1 Role and Purpose of Crowdsensing and Crowdsourcing, Namely Geofences and Location-Based Notifications, in Smart Cities
4.2 Strategies and Tools that have Already Been Developed and Used to Implement and Integrate Mechanisms of Crowdsensing and Crowdsourcing in Smart Cities
4.3 Improvement of Citizens' Lives in a Smart City Through Crowdsensing and Strategies to Make Citizens Active Contributors
5 Conclusion
References
Understanding Urban Mobility Habits and Their Influencing Factors on a University Campus in Argentina
1 Introduction
2 Conceptual Framework
3 Methodology
4 Results and Discussion
4.1 Description of the Case Study
4.2 Descriptive Analysis of Urban Mobility of the University Community
4.3 Analysis of Factor Influencing Urban Mobility Decisions
5 Final Remarks
References
A Conversational Agent for Smart Schooling A Case Study on K-12 Dropout Risk Assessment
1 Introduction
2 The Case Study Report
2.1 Project Description
2.2 Component View
2.3 Education Intelligence
2.4 Conversational Agent
2.5 Realworld Experiments
3 Results and Discussion
4 Conclusion
References
Evaluation of Smart Mobility Indicators in Latin-American Cities
1 Introduction
2 Evaluation of Smart Mobility Models
3 Mobility Challenges in Latin America
4 Methodology
5 Results and Analysis
6 Conclusions and Future Work
References
Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation
1 Introduction
2 Models for ICT Adoption and Appropriation
3 Consensus of Characteristics for ICT Adoption and Appropriation
3.1 Interpretative Structural Modeling (ISM)
3.2 Model Review and Consensus Building
3.3 Expert Validation—Accessibility Matrix
3.4 Relationship Graph
4 Conclusions
References
Trends of Artificial Intelligence and Blockchain in New Public Management
1 Introduction
2 Research Methodology
2.1 Definition of Research Question
2.2 Search
2.3 Study Selection
2.4 Data Extraction
3 Results
3.1 Publication and Geographical Distribution
3.2 Summary of Contributions
3.3 Contributions by Authors and Characteristics for Blockchain and Artificial Intelligence and New Public Management
4 Discussion
5 Conclusions
References
A Novel Approach for Smart Mobility Roadmap in Smart Cities
1 Smart Cities Overview
1.1 Smart Mobility as a Driver in the City
2 Importance of Key Performance Indicators to Build Smart Mobility in Smart Cities
2.1 Smart Mobility KPIs by the International Organization for Standardization
2.2 Smart Mobility KPIs by International Telecommunication Union
2.3 Smart Mobility KPIs in Colombia
3 Smart Mobility KPIs Comparison
4 Smart Mobility Roadmap Approach
5 Conclusions
References
Exploring the Use of Digital Twin in Smart Healthcare: A Case Study of Dengue Epidemic Control and Prevention
1 Introduction
2 Background
2.1 Digital Twin
2.2 The Modelling and Simulation of Complex Systems
2.3 Relate Work
3 Methodology
3.1 Materials and Methods
4 Results
4.1 System Dynamics Model for Preventing Dengue
4.2 Smart Health Platform Architecture
4.3 Implementation Progress
5 Discussion and future work
References
Industry 4.0 Smart Supply Chain Data Protection Using Block Chain
1 Introduction
2 Background
2.1 Blockchain
2.2 Types of Blockchain
2.3 Different Types of Consensus Algorithms
2.4 Supply-Chain Relation with Blockchain
3 Methods
3.1 System Design
3.2 Hyperledger Fabric Binaries (Tools)
3.3 Orderers
3.4 Peers
4 Channels
5 Results
5.1 Generating the Crypto Material and Lauching Orderers
5.2 Launching Peers
5.3 Joining Channels
6 Conclusion
References
Methodological Model for the Solution of Periodic Customer Scheduling in Routing Problems
1 Introduction
2 Related Works
3 Problem Formulation
3.1 Variable Definition
4 Methodology
5 Results
5.1 Node Distance Reduction (dij)
5.2 Solving the Visit Scheduling Problem
6 Conclusions
References
Development of an Information System for the Monitoring of Physiological Variables and Information Storage of Neonates on Domiciliary Oxygen
1 Introduction
2 Methodology
2.1 Software Development
2.2 Platform Development
2.3 Results
3 Discussion
4 Conclusions
References
Principal Component Analysis for Knowledge Transfer in the Social Structure Reconstruction Program in Post-conflict Zones in Colombia (Chocó, Sucre and Caldas)
1 Introduction
2 Theoretical Framework
2.1 Knowledge Transfer and Peace Processes
3 Methodology
4 Results
5 Discussions
6 Conclusions
References
Epileptic Seizure Prediction Methods Using Machine Learning and Deep Learning Models
1 Introduction
2 Material and Methods
2.1 Dataset
2.2 Model
2.3 Metrics
3 Results
4 Conclusions
References
Towards Energy Efficiency in Microgrids for Smart Sustainable Cities
1 Introduction
2 Related Work
3 Monitoring Energy Data in Microgrids Using Smart Meters
4 Case Study
5 Conclusions
References
Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion in Medium-Sized Latin American City Streets
1 Introduction
2 Study Area
3 Methodology and Data Analysis
4 Results and Discussions
4.1 Vehicular Flow of the Study Area and Its Segmentation
4.2 Meteorology and Modeling in the Study Area
5 Conclusions
References
Evaluation of Feature Descriptors for Scene Classification
1 Introduction
2 Feature Representations
2.1 Descriptors of Global Characteristics
2.2 Local Feature Extraction
3 Experimental Evaluation and Results
3.1 Database
3.2 Experimental
3.3 Discussion
4 Conclusion
References
Influence of Segmentation Schemes on the Interpretability of Functional Connectivity in Mild Cognitive Impairment
1 Introduction
2 Methods
2.1 EEG Data Preprocessing
2.2 EEG Change Point Detection Analysis
2.3 Functional Connectivity
3 Results and Discussions
3.1 EEG Segmentation Method Signal Analysis
3.2 Functional Connectivity
4 Conclusions
References
The Content Based Misinformation Detection for Gujarati Language
1 Introduction
2 Literature Review
3 Proposed Architecture
4 Experimental Results
4.1 Data Collection
4.2 Preprocessing
4.3 Major Findings
5 Conclusion
References
Analysis of Indicators in the Urban Distribution of Food in Megacities Within the Framework of Sustainability
1 Introduction
2 Methodology
2.1 Data Selected for Analysis
2.2 Sustainability Indicator Model
3 Results
3.1 Descriptive Analysis of Variables
3.2 Multivariate Model
3.3 Correlations and Dendogram
3.4 Aggregate Sustainability Indicator
4 Conclusions
References
Passenger Forecasting as a Mobility Project for Smart Cities in Manizales, Colombia
1 Introduction
2 Methodology
2.1 Dataset
2.2 Data Preprocessing
2.3 Model Architecture
2.4 Hyperparameters
2.5 Recursive Test
2.6 Metrics
2.7 Integration with Dashboard
3 Results
3.1 Descriptive Results
3.2 Predictive Results
4 Discussion
5 Conclusion
5.1 Future Work
References
CO2 Emissions Curtailment from the Usage of Electric Vehicles in Sal Island
1 Introduction
2 Methodology
2.1 Vehicle Fleet Emissions
2.2 Electric Energy Consumption
3 Case Study
3.1 Environmental Context
3.2 Electricity Sources Characterisation
4 Results and Discussion
5 Conclusion
References
Smart, Rather Than Intelligent: A Study on Signifiers and Their Implications for Cities Management
1 Introduction
2 Intelligent vs. Smart
3 Innovation, Development & Smart Cities
4 Talent and Adaptation
5 Talent and Adaptation
6 Conclusions and Recommendations for Future Studies
References
International Workshop on Climate Change, Tourism and Education (CCTE)
Sustainable Historic Urban Landscape: Valuation Tools
1 Introduction
2 Historic Urban Landscape
3 Regulatory Guidance
4 Participation and Empowerment
5 Heritage Communities
6 Cultural Dimension of Development
7 The Valuation of Urban Heritage from the Smart Cities Approach
8 Conclusions
References
Do Emotions Matter? Reviewing the Last Generation of Studies on Climate Change Communication and Tourist Behaviour
1 Introduction
2 Literature Review
2.1 Approaches to Measure Emotions and Tourism
2.2 Climate Change Perception and Action in Tourism
2.3 The Role of Communication in Climate Change Awareness Raising
3 Method
4 Results
5 Conclusions
References
The Training of Territorial Enhancers for the Valorization and Management of Biocultural Assets
1 Introduction
2 The SUS-TER Project and the Training of Territorial Enhancers
3 The University of Caldas in the SUS-TER Project: Impact of the Training of Territorial Enhancers
3.1 Institutional Strategy
3.2 Student Profile
3.3 Territorial Dynamization Initiatives Developed by Students
4 Evaluation and Learning
5 Final Considerations
References
Rural Tourism Management as a Route to a Smart Territory Within the Framework of the Triple Transition. Opportunities from a Colombian Rural Territory
1 Introduction
2 Tourism as an Economic Activity in Colombia
3 Particularities of the Territory and its Tourism Potential
4 Rural Tourism and Smart Territories
5 What Makes Territories Intelligent?
6 Conclusions
References
International Workshop on Intelligent Systems Applied In Adaptive Smart Areas (ISAIAS)
Monitoring System for Detecting Non-inclusive Situations in Smart Cities
1 Introduction
2 Related Work
2.1 Detection of Non-inclusive Needs in Smart Cities
2.2 Deep Learning Algorithms for Detecting Non-inclusive Situations
3 Deep Learning Proposal
3.1 OpenCV to Project the Centroid
3.2 YoloV5 to Detect Persons in Situations of Non-inclusivity
4 Case Study
4.1 Experimental Set-Up
4.2 Results
5 Conclusions and Future Work
References
Evaluation of XAI Models for Interpretation of Deep Learning Techniques' Results in Automated Plant Disease Diagnosis
1 Introduction
2 Background
2.1 Challenges in Plant Disease Detection
2.2 Convolutional Neural Networks
2.3 Explainable Artificial Intelligence
2.4 Convolutional Neural Networks for Plant Disease Detection
2.5 Challenges in Plant Disease Detection
2.6 Disease Detection in Farming Environments
2.7 Conclusion
3 Case Study
3.1 General Description of the Experiment
3.2 Experimental Results
4 Conclusions
References
Edge Service Allocation Based on Clustering Techniques
1 Introduction
2 Related Works
3 Proposed Framework
3.1 Information Layer
3.2 Filter Layer
3.3 Decision Layer
4 Allocation Example
5 Conclusions and Future Work
References
A Brief Review of Explainable Artificial Intelligence (XAI) Techniques
1 Introduction
2 What is Explainable AI?
3 Review of Techniques
3.1 Ad-Hoc Techniques
3.2 Post-hoc Techniques
4 Practical Application
5 Conclusions
References
Dynamic Transfer Point Allocation for Rural Demand-Responsive Mobility
1 Introduction
2 Related Work
3 Problem Formulation
3.1 Optimizing DRT Systems Through Dynamic Transfer Point Allocation
3.2 Formulation
3.3 Modeling Constraints
4 Discussion
4.1 The Effect of Operational Cost Optimization on Passenger Experience
4.2 Adoption Rate in Flexible Transportation Systems
5 Conclusions
References
Doctoral Consortium
Convolutional Neural Network for DDoS Detection
1 Introduction
2 Related Work
3 Materials and Method
3.1 Processing of the Dataset
4 Results
5 Conclusions
References
Trends and Applications of Heuristic Algorithms in Transportation Logistics
1 Introduction
2 Heuristic Algorithms in Transportation Logistics
3 Trends in Transport Logistics
4 Maritime Route Optimization
5 Conclusions
References
The Effects of Eco and Smart Policies: A Social Justice Perspective
1 Introduction
2 The EU Eco-smart Trends
3 Effects of Eco-smart Policies
3.1 Social Effects on Housing: Ecological Gentrification
3.2 Social Effects on Labour: Job Displacement
4 Counter Strategies
5 Conclusion
References
Author Index
Recommend Papers

Trends in Sustainable Smart Cities and Territories
 9783031369575, 3031369572

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Lecture Notes in Networks and Systems 732

Luis Fernando Castillo Ossa · Gustavo Isaza · Óscar Cardona · Omar Danilo Castrillón · Juan Manuel Corchado Rodriguez · Fernando De la Prieta Pintado   Editors

Trends in Sustainable Smart Cities and Territories

Lecture Notes in Networks and Systems Volume 732

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]).

Luis Fernando Castillo Ossa · Gustavo Isaza · Óscar Cardona · Omar Danilo Castrillón · Juan Manuel Corchado Rodriguez · Fernando De la Prieta Pintado Editors

Trends in Sustainable Smart Cities and Territories

Editors Luis Fernando Castillo Ossa Department of Systems and Information Technology Universidad de Caldas Manizales, Colombia

Gustavo Isaza Department of Systems and Information Technology Universidad de Caldas Manizales, Colombia

Óscar Cardona Department of Electronics and Industrial Automation Universidad Autónoma de Manizales Manizales, Colombia

Omar Danilo Castrillón Department of Industrial Engineering Universidad Nacional de Colombia Sede Manizales Manizales, Colombia

Juan Manuel Corchado Rodriguez BISITE Research Group, Edificio Multiusos I+D+i Universidad de Salamanca Salamanca, Spain

Fernando De la Prieta Pintado BISITE Research Group, Edificio Multiusos I+D+i Universidad de Salamanca Salamanca, Spain

ISSN 2367-3370 ISSN 2367-3389 (electronic) Lecture Notes in Networks and Systems ISBN 978-3-031-36956-8 ISBN 978-3-031-36957-5 (eBook) https://doi.org/10.1007/978-3-031-36957-5 © 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

Preface

The world’s population is expected to reach 9.7 billion in 2050. By then, two-thirds of the population will live in urban environments. Critical social and ecological challenges that cities will face include urban violence, inequality, discrimination, unemployment, poverty, unsustainable energy and water use, epidemics, pollution, environmental degradation and increased risks of natural disasters. The concept of smart cities, which emerged in the early 2000s, attempts to solve these challenges by implementing information and communication technologies. The initial concept of smart cities focused on the modernisation of megacities. However, most so-called smart cities are just cities with several ‘smart’ projects. The most promising trend is the creation of smart territories, defined as small hi-tech towns, districts or satellite towns near megacities. With the current availability of an enormous amount of data, the challenge is to identify intelligent and adaptive ways of combining the information to create valuable knowledge. Sensorisation plays a fundamental role in data collection, which, once analysed on IoT and smart city platforms, can be used to make multiple decisions regarding governance and resource consumption optimisation. The International Conference on Sustainable Smart Cities and Territories (SSCt’23) is an open symposium that brings together researchers and developers from academia and industry to present and discuss the latest scientific and technical advances in the fields of smart cities, smart territories and, in general terms, smart areas. It promotes an environment for discussion on how techniques, methods and tools help system designers accomplish the transition from the current cities toward those we need in a changing world. The SSCt’23 technical programme includes a main technical track, doctoral consortium and two related workshops such as Climate Change, Tourism and Education (CCTE) and Intelligent Systems Applied in Adaptive Smart Areas (ISAIAS). All papers underwent a peer-review selection, assessed by three reviewers from an international panel of about 144 members from 16 countries. On average, the quality of submissions was good, from 40 submissions received, 30 were selected. It presents high quality and diversity, bringing together authors from many countries (Argentina, Cabo Verde, Cameroon, Colombia, Egypt, India, Panama, Paraguay, Peru, Portugal, Qatar, Senegal, Spain, Switzerland and Thailand) and various subfields in smart v

vi

Preface

cities and territories. Therefore, this event will strongly promote interaction among researchers from international research groups working in diverse fields. The scientific content will be innovative and help improve the valuable work the participants are carrying out. This symposium is organised by the University of Caldas (Colombia) with the collaboration of the University of Salamanca (Spain) and the AIR Institute (Spain). We want to thank all the contributing authors, the members of the Programme Committee, the sponsors Gobierno de Caldas, People Contact, Alcaldía de Manizales, Universidad Nacional de Colombia, Universidad Autónoma de Manizales, Universidad de Manizales, Visit Manizales, and Camara de Comercio de Manizales, BIOS, PROCOLOMBIA, and the Local Organisation members for their valuable work, which has been essential for the success of SSCt’23. Manizales, Colombia Manizales, Colombia Manizales, Colombia Manizales, Colombia Salamanca, Spain Salamanca, Spain

Luis Fernando Castillo Ossa Gustavo Isaza Oscar Cardona Omar Danilo Castrillón Juan Manuel Corchado Rodriguez Fernando De la Prieta Pintado

Organization

Local Committee • Luis Fernando Castillo Ossa (Chair), Universidad de Caldas, Universidad Nacional de Colombia Sede Manizales, Colombia • Marcelo López Trujillo (Industrial Chair), Universidad de Caldas, Universidad Nacional de Colombia Sede Manizales, Colombia • Sandra Victoria Hurtado, Universidad de Caldas, Colombia • Oscar Hernán Franco, Universidad de Caldas, Colombia • Gustavo Isaza, Universidad de Caldas, Colombia • Jeferson Arango Lopez, Universidad de Caldas, Colombia

Partners Committee • Diego Hernando Ceballos (Gerente de People Contact), Manizales, Colombia • Jorge Alberto Jaramillo Garzon, BIOS • Juan Felipe Jaramillo, Secretaria TIC y Competitividad—Alcaldía de Manizales, Colombia • Nicolas Llano, Representante Secretaría De Desarrollo, Empleo e innovación— Gobernación de Caldas, Colombia • Santiago Ruiz Herrera, Decano de la Facultad de Ingeniería y Arquitectura Universidad Nacional de Colombia Sede Manizales, Colombia • Omar Danilo Castrillón, Departamento Ingeniería Industrial UNAL Sede Manizales, Colombia • Oscar Cardona, Universidad Autónoma de Manizales, Colombia • Santiago Murillo, Universidad Autónoma de Manizales, Colombia • Luis Carlos Correa, Universidad de Manizales, Colombia

vii

viii

Organization

• Johana Yepez, Secretaria TIC y Competitividad—Alcaldía de Manizales, Colombia • Maria Idally Buitrago, Cámara y Comercio de Manizales, Colombia

Workshops Chairs • Fernando de la Prieta, University of Salamanca, Spain • Oscar Hernán Franco, Universidad de Caldas, Colombia • Saber Trabelsi, Science Program, Texas A&M University, Qatar

Doctoral Consortium Chairs • Sara Rodríguez, University of Salamanca, Spain • Gustavo Adolfo Isaza, Universidad de Caldas, Colombia • Omar Danilo Castrillón, Universidad Nacional de Colombia Sede Manizales, Colombia

Programme Committee • • • • • • • • • • • • • • • • •

Juan Manuel Corchado (co-chair), University of Salamanca, Spain Sigeru Omatu (co-chair), Hiroshima University, Japan Mahmoud Abbasi, University of Salamanca, Spain Juan M. Alberola, Universitat Politècnica de València, Spain Ana Almeida, ISEP-IPP, Portugal Cesar Analide, University of Minho, Portugal Luis Antunes, University of Lisbon, Portugal Jeferson Arango Lopez, Universidad de Caldas, Colombia Iván Bernabé, Rey Juan Carlos University of Madrid, Spain Holger Billhardt, Universidad Rey Juan Carlos, Spain Luis Blazquez, University of Salamanca, Spain Vicent Botti, Universitat Politècnica de València, Spain Carlos Carrascosa, GTI-IA DSIC Universidad Politecnica de Valencia, Spain Eduardo Carrillo, Universidad Autónoma de Bucaramanga, Colombia Fabio Cassano, Università degli Studi di Bari Aldo Moro, Italy Luis Fernando Castillo Ossa, Universidad de Caldas, Colombia Omar Danilo Castrillon Gomez, Universidad Nacional de Colombia Sede Manizales, Colombia • Pablo Chamoso, University of Salamanca, Spain • Cesar A. Collazos, University of Cauca, Colombia

Organization

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

ix

Rafael Corchuelo, University of Seville, Spain Giovanni De Gasperis, DISIM, Università degli Studi dell’Aquila, Italy Luis De La Fuente Valentín, Universidad Internacional de La Rioja, Spain Fernando De La Prieta, University of Salamanca, Spain Javier De León Ledesma, University of Las Palmas de Gran Canaria, Spain Raffaele Dell’Aversana, Università “D’Annunzio” di Chieti-Pescara, Italy Yves Demazeau, CNRS—LIG, France Tania Di Mascio, DISIM, University of L’Aquila, Italy Enrique Diaz-Plaza, University of Salamanca, Spain Nestor Duque, Universidad Nacional, Colombia Dalila Alves Durães, Universidade do Minho, Portugal Liliana Durón, University of Salamanca, Spain Ponciano Jorge Escamilla-Ambrosio, CIC-IPN, Mexico Fernando Escobar, University of Minho, Portugal Rino Falcone, Institute of Cognitive Sciences and Technologies-CNR, Italy Florentino Fdez-Riverola, University of Vigo, Spain Alberto Fernandez, University Rey Juan Carlos, Spain Antonio Fernández-Caballero, Universidad de Castilla-La Mancha, Spain Angela Ferreira, Polytechnic Institute of Bragança, Portugal Ruben Fuentes-Fernandez, Universidad Complutense de Madrid, Spain Oscar Garcia, UNIR, Spain Rosella Gennari, Free U. of Bozen-Bolzano, Italy Faber Giraldo, Universidad del Quindío, Colombia Sylvain Giroux, Université de Sherbrooke, Canada Jorge Gomez-Sanz, Universidad Complutense de Madrid, Spain Alfonso González Briones, University of Salamanca, Spain Carina Gonzalez-González, Universidad de La Laguna, Spain Ivan Gutierrez, TECNALIA, Spain Gustavo Isaza, University of Caldas, Colombia Vicente Julian, Universitat Politècnica de València, Spain Marcelo Karanik, Rey Juan Carlos University of Madrid, Spain Oscar Lage, TECNALIA, Spain Yen Elizabeth Lam González, Institute of Tourism and Sustainable Economic Development (TiDES), Spain Sandeep Langar, The University of Texas at San Antonio, United States Paulo Leitao, Polythecnic Institute of Braganca, Portugal Tiancheng Li, Northwestern Polytechnical University, China Raúl López Blanco, Universidad de Salamanca, Spain José Machado, University of Minho, Portugal Gonçalo Marques, Polytechnic of Coimbra, Portugal Sergio Márquez, BISITE, Spain Goreti Marreiros, ISEP/IPP-GECAD, Portugal Angel Martin Del Rey, Department of Applied Mathematics, Universidad de Salamanca, Spain Eric Matson, Purdue University, United States

x

Organization

• • • • • • • • • • • • • • • • • • • •

Carlos Meza Benavides, Anhalt University of Applied Sciences, Germany Jose M. Molina, Universidad Carlos III de Madrid, Spain Pablo Monzon, Udelar, Uruguay Juan Jose Morillas Guerrero, Universidad Politécnica de Madrid, Spain Paulo Novais, University of Minho, Portugal Sascha Ossowski, University Rey Juan Carlos, Spain Javier Parra, University of Salamanca, Spain Juan Pavón, Universidad Complutense de Madrid, Spain Pawel Pawlewski, Poznan University of Technology, Poland Hugo Peixoto, University of Minho, Portugal Antonio Pereira, Escola Superior de Tecnologia e Gestão do IPLeiria, Portugal António Pinto, ESTG, P.Porto, Portugal Pedro Pinto, Instituto Politécnico de Viana do Castelo, Portugal Tiago Pinto, ISEP, Portugal Francisco Pinto-Santos, BISITE Research Group, Spain Marta Plaza-Hernández, BISITE Research Group, University of Salamanca, Spain Jose-Luis Poza-Luján, Universitat Politècnica de València, Spain Isabel Praça, GECAD/ISEP, Portugal Javier Prieto, University of Salamanca, Spain Araceli Queiruga-Dios, Department of Applied Mathematics, Universidad de Salamanca, Spain Joao Ramos, School of Management and Technology—Polytechnic Institute of Porto (IPP), Portugal Miguel Rebollo, Universitat Politècnica de València, Spain Bernardo Rivera Sánchez, Universidad de Caldas, Colombia Sara Rodríguez, University of Salamanca, Spain Ricardo Santos, ESTG/IPP, Portugal Ichiro Satoh, National Institute of Informatics, Japan Fernando Silva, School of Technology and Management, Computer Science and Communications Research Centre, Polytechnic of Leiria, Portugal Reinel Tabares, Universidad Nacional de Colombia, Colombia Andrei Tchernykh, CICESE Research Center, Mexico Zita Vale, GECAD—ISEP/IPP, Portugal Rafael Valencia-Garcia, Departamento de Informática y Sistemas. Universidad de Murcia, Spain José Ramón Villar, University of Oviedo, Spain Pierpaolo Vittorini, University of L’Aquila—Department of Life, Health, and Environmental Sciences, Italy Jian-Guo Zhang, London South Bank University, United Kingdom

• • • • • • • • • • • • • •

Organization

Sponsors

xi

xii

Organization

International Workshop on Climate Change, Tourism and Education (CCTE) Scope Research on the interrelationships of weather, climate and climate change with tourism and recreation has developed exponentially over the past decades. These investigations have served to improve climate services and provide recommendations on alternative pathways for climate change adaptation and mitigation. There is a significant progress in the study of vulnerability and risks affecting the tourism activity, as well as the contribution of tourism to both, GHG emissions and mitigation and carbon off-setting programmes. There is also a consensus that large-scale climatic projections do not derive valid statements at local level, as climate change impacts are heterogeneous across nations and communities. Also, adaptation to a changing climate is fundamentally a local issue, and local involvement, awareness, participation, and ownership are a central precondition for a successful implementation of actions. But the tourism sector and the academy still lack capacity to provide the massive information requirements to deal with this looming climatic crisis we are currently living (Leon et al., 2021). Beyond COVID-19, global tourism will need to respond to the physical, social and economic changes that climate change will drive. Although climate change and COVID-19 crises differ in many fundamental ways, including the speed at which they develop, the health crisis caused by COVID-19 has led to important research challenges and questions to be addressed in the tourism sector. For instance, the analysis of the effectiveness of trustworthiness, openness, negative, optimism, or valence, and the communication channel to trigger tourist emotions and encourage them in more environmentally responsible behaviour. Finally, environmental education is another powerful instrument to fight against climate change, strengthening the capacity and willingness of people, enterprises and communities to lead significant changes which help to avoid the worst possible scenarios of the climate crisis. Much of what has been written in this context is still conceptual and there are very few empirical insights. This Workshop aims to bring together the scholars, researchers, students and experts from around the world and different disciplines to explore recent developments in tourism climatology, environmental studies, climate change, tourism management, climate and tourism policy, communication and psychology studies, among others. Participants are invited to share their applied research on the following suggested topics, although any other relevant subject will be welcomed. • COVID-19 and climate change crises, dissimilarities and lessons learned with implications to tourism. • Climate change communication challenges to accelerate a behavioural change of tourists, firms and other tourism stakeholders. • Climate Change, emotions and tourist behaviour.

Organization

xiii

• Bottom-up approaches for climate change policy/education design in the tourism field. • Applied analysis to the impacts of climate change on tourist destinations. • Climate change inclusion in tourism education and research. • Weather sensitivity and climate change perceptions of tourists. • Approaches for the study of climate-related risks and vulnerability for tourism.

Acknowledgment This workshop is a product of the Erasmus + Project ‘Strengthening research, innovation and knowledge transfer on Climate Change & Tourism in Higher Education Institutions in Latin America’. University of Caldas and University of Alicante are beneficiary partners of this project, and the University of Las Palmas de Gran Canaria, coordinator, with the help of Obreal Global.

Organizing Committee • Bernardo Rivera Sánchez (President), Director del Grupo de Investigación en Análisis de Sistemas de Producción Agropecuaria (ASPA), University of Caldas, Colombia • Javier de León Ledesma, Institute of Tourism and Sustainable Economic Development (TiDES), UNESCO Chair in Tourism and Sustainable Economic Development, University of Las Palmas de Gran Canaria • Jimena Estrella Orrego, Faculty of Agricultural Sciences, National University of Cuyo (Argentina), Advisor for Innovation Obreal Global, Barcelona

xiv

Organization

Scientific Committee • Yen E. Lam-González (President), Institute of Tourism and Sustainable Economic Development (TiDES) • Andrés Felipe Betancourth López, Governance and Territorial Development Area for Latin America and the Caribbean, Colombia • Carmelo J. León González, Institute of Tourism and Sustainable Economic Development (TiDES), UNESCO Chair in Tourism and Sustainable Economic Development, University of Las Palmas de Gran Canaria, Spain • Oana Mãdãlina Driha, Institute of International Economics, University of Alicante, Spain • Lucellys Sierra-Márquez, Doctoral Program in Environmental Toxicology, Environmental and Computacional Chemistry Group, University of Cartagena, Colombia • Martín Alberto Rodríguez Brindis, Research and Postgraduate Center, Anahuac Oaxaca University, Mexico

Organization

xv

International Workshop on Intelligent Systems Applied In Adaptive Smart Areas (ISAIAS) Scope The purpose of this workshop is to explore technological solutions for Adaptive Smart Areas, which are geo-located locations that require extensive sensorization to adapt to significant changes in environmental conditions. The proposed solutions should facilitate decision-making for independent entities and promote collaboration and coordination among them to enhance available resources and increase efficiency in various areas such as cities, buildings, villages, farms and forests. Adaptive Smart Areas integrate energy efficiency, sustainable mobility, environmental protection and economic sustainability to provide intelligent agents with unlimited opportunities to display their abilities to react, plan, learn and interact in an intelligent and human-like manner. The workshop focuses on the use of intelligent systems such as MAS (Multiagent systems) to provide intelligence to Adaptive Smart Areas. Papers on the use of agents in Adaptive Smart Areas addressing issues related to smart architectures, simulations, intelligent infrastructure, smart transport, robotics, open data, and specific methodological and technological issues related to the real deployment of agents are welcome. Topics that could be relevant for the workshop include specially applications, but also theoretical approaches, based (but not limited) on: • • • • • • • • • • • • • • • • • •

Smart health and emergency management Smart environments Smart education Smart health Smart mobility and transportation ML-enabled predictive and analytic solutions Intelligence applications and crop monitoring Smart city modelling Smart city simulation Smart farming Intelligent infrastructures Sensors and actuators Ethical and legal issues in Adaptive Smart Areas Intelligent vehicles Cloud and edge assisted applications Agricultural robotics Cost-efficient wireless sensor nodes, network architecture and implementation New environment sensor technologies

xvi

Organization

Programme Committee Chairs • • • • • •

Iván Bernabé, Rey Juan Carlos University of Madrid, Spain Carlos Carrascosa, Universitat Politècnica de València, Spain Fernando de la Prieta, University of Salamanca, Spain Vicente Julian, Universitat Politècnica de València, Spain Marcelo Karanik, Rey Juan Carlos University of Madrid, Spain Sara Rodríguez, University of Salamanca, Spain

Scientific Committee • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Marcelo Albornoz, CONICET, Argentina Ricardo S. Alonso Rincón, University of Salamanca, Spain Joaquín Arias, Rey Juan Carlos University of Madrid, Spain Holger Billhardt, Rey Juan Carlos University of Madrid, Spain Pablo Chamoso Santos, University of Salamanca, Spain Juan Manuel Corchado Rodríguez, University of Salamanca and AIR institute, Spain Alberto Fernández, Rey Juan Carlos University of Madrid, Spain Ana Belén Gil González, University of Salamanca, Spain Adriana Giret, Universitat Politècnica de València, Spain Diego Godoy, Universidad Gastón Dachary, Argentina Angélica González Arrieta, University of Salamanca, Spain Alfonso González Briones, University of Salamanca, Spain Guillermo Hernández González, AIR institute, Spain Jaume Jordan, Universitat Politècnica de València, Spain David Luis La Red Martínez, Universidad Tecnológica Nacional, Argentina Marin Lujak, Rey Juan Carlos University of Madrid, Spain Cedric Marco, Universitat Politècnica de València, Spain Claudio Roberto Marquetto Mauricio, Universidade Estadual do Oeste do Paraná, Brasil Sergio Marquez, University of Salamanca, Spain Pasqual Martí, Universitat Politècnica de València, Spain Yeray Mezquita Martín, University of Salamanca, Spain José Arturo Mora Soto, Universidad de Celaya, México Sascha Ossowski, Rey Juan Carlos University of Madrid, Spain Javier Palanca, Universitat Politècnica de València, Spain Javier Parra, University of Salamanca, Spain Marta Plaza Hernández, University of Salamanca, Spain Javier Prieto Tejedor, University of Salamanca and AIR institute, Spain Miguel Rebollo, Universitat Politècnica de València, Spain Jaime Rincon, Universitat Politècnica de València, Spain Andrés Terrasa, Universitat Politècnica de València, Spain

Contents

Main Track Motorized Mobility on a Latin American University Campus: A Preliminary Study Focused on Sustainability . . . . . . . . . . . . . . . . . . . . . . . Yamila S. Grassi, Mónica F. Díaz, Gabriela Pesce, Florencia Pedroni, María Andrea Rivero, and Héctor G. Chiacchiarini

3

Smart Urban Governance Through Geoinformation: The Importance of Geoportals for City Interoperability . . . . . . . . . . . . . . . . . . . Luiz Ugeda and Isabel Celeste Fonseca

15

Digital Twin Application Methodology for the Improvement of Production and Service Systems. Application to Waste Management Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jhonathan Mauricio Vargas Barbosa, Omar Danilo Castrillon Gomez, and Jaime Alberto Giraldo García

25

Adaptive and Intelligent Edge Computing Based Building Energy Management System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sergio Márquez-Sánchez, Sergio Alonso-Rollán, Francisco Pinto-Santos, Aiman Erbad, Muhammad Hanan Abdul Ibrar, Javier Hernandez Fernandez, Mahdi Houchati, and Juan Manuel Corchado

37

A Novel Aging-Based Proof of Stake Consensus Mechanism . . . . . . . . . . . Mahmoud Abbasi, Javier Prieto, Marta Plaza-Hernández, and Juan Manuel Corchado

49

Manizales, Smart and Sustainable Tourist Destination . . . . . . . . . . . . . . . . Luis Carlos Correa-Ortiz, Catalina Guevara-Giraldo, and Elizabeth Chaparro Cañola

62

xvii

xviii

Contents

Development of an IoT Network for Urban Orchards in High Vulnerability Areas in Colombia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Juan M. Núñez V, Juan Manuel Corchado, and Diana M. Giraldo DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gustavo Isaza, Fabian Ramirez, Néstor Duque, Jeferson Arango Lopez, and José Montes Smart Cities Using Crowdsensing and Geoferenced Notifications . . . . . . . Rui Miranda, Eduarda Ribeiro, Dalila Durães, Hugo Peixoto, Ricardo Machado, António Abelha, and José Machado

75

87

97

Understanding Urban Mobility Habits and Their Influencing Factors on a University Campus in Argentina . . . . . . . . . . . . . . . . . . . . . . . . 111 Gabriela Pesce, Florencia Pedroni, María Andrea Rivero, Héctor G. Chiacchiarini, Yamila S. Grassi, and Mónica F. Díaz A Conversational Agent for Smart Schooling A Case Study on K-12 Dropout Risk Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Renata Magalhães, Bruno Veloso, Francisco S. Marcondes, Henrique Lima, Dalila Durães, and Paulo Novais Evaluation of Smart Mobility Indicators in Latin-American Cities . . . . . 135 Eladio E. Martinez Toro, Elian Krut Yalil, and Vit Bubak Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Daniel Montes Agudelo, Jheimer Julián Sepúlveda López, and Luz Arabany Ramírez Castañeda Trends of Artificial Intelligence and Blockchain in New Public Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 Jhon Wilder Sanchez-Obando and Luis Fernando Castillo Ossa A Novel Approach for Smart Mobility Roadmap in Smart Cities . . . . . . . 170 Tatiana Zona-Ortiz and Gustavo Guzman Exploring the Use of Digital Twin in Smart Healthcare: A Case Study of Dengue Epidemic Control and Prevention . . . . . . . . . . . . . . . . . . . 183 Andres Rey Piedrahita, Jenniffer Castellanos-Garzón, Julián Eduardo Betancur, Marco Tulio Canizales, Juan Sebastián Henao-Agudelo, Luis Alberto Rivera Martinez, and Sebastian Lopez-Mejia Industry 4.0 Smart Supply Chain Data Protection Using Block Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 Mohamed Ayman Abu El Magd, Nada Sharaf, and Maggie Mashaly

Contents

xix

Methodological Model for the Solution of Periodic Customer Scheduling in Routing Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Restrepo Franco Alejandra María, Valencia Rodriguez Orlando, Toro Ocampo Eliana Mirledy, Bravo Ortíz Mario Alejandro, Cardona Ramirez Nicolas, Orjuela Paez Cristian Camilo, and Valencia Díaz Mario Andrés Development of an Information System for the Monitoring of Physiological Variables and Information Storage of Neonates on Domiciliary Oxygen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 J. L. Amaya-Carrascal, C. A. Vera-Betancourt, C. Marquez-Narvaez, S. Murillo-Rendón, I. Echeverri-Ocampo, D. Henao-Díaz, Belarmino Segura-Giraldo, and C. Salgado-Jiménez Principal Component Analysis for Knowledge Transfer in the Social Structure Reconstruction Program in Post-conflict Zones in Colombia (Chocó, Sucre and Caldas) . . . . . . . . . . . . . . . . . . . . . . . 231 Marcelo López, Germán Gómez, and Carlos Marulanda Epileptic Seizure Prediction Methods Using Machine Learning and Deep Learning Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 Maria Alejandra Patiño-Claros, Sergio Alejandro Holguin-García, Alvaro Eduardo Daza-Chica, Reinel Tabares-Soto, and Mario Alejandro Bravo-Ortiz Towards Energy Efficiency in Microgrids for Smart Sustainable Cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254 V. Isanbaev, R. Baños, C. Gil, M. M. Gil, F. Martínez, and A. Alcayde Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion in Medium-Sized Latin American City Streets . . . . 266 Yamila S. Grassi and Mónica F. Díaz Evaluation of Feature Descriptors for Scene Classification . . . . . . . . . . . . . 277 Luis Hernando Ríos González, Sebastián López Flórez, Alfonso González-Briones, and Fernando de la Prieta Influence of Segmentation Schemes on the Interpretability of Functional Connectivity in Mild Cognitive Impairment . . . . . . . . . . . . . 289 Isabel Echeverri-Ocampo, Karen Ardila, José Molina-Mateo, Jorge Iván Padilla-Buriticá, Belarmino Segura-Giraldo, Hector Carceller, Ernesto A. Barceló-Martinez, and Maria de la Iglesia-Vaya The Content Based Misinformation Detection for Gujarati Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298 Uttam Chauhan, Vinay Sheth, Vishvesh Trivedi, Chintan Bhatt, and Juan Manuel Corchado

xx

Contents

Analysis of Indicators in the Urban Distribution of Food in Megacities Within the Framework of Sustainability . . . . . . . . . . . . . . . . 310 Leila Nayibe Ramírez Castañeda, Sonia Lucila Meneses, Astrid Altamar, and Edwin Bulla Passenger Forecasting as a Mobility Project for Smart Cities in Manizales, Colombia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 Daniel Arias-Garzón, Johan Sebastian Piña Duran, Diego Hernando Ceballos López, and Reinel Tabares-Soto CO2 Emissions Curtailment from the Usage of Electric Vehicles in Sal Island . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332 Randi Graça and Ângela Ferreira Smart, Rather Than Intelligent: A Study on Signifiers and Their Implications for Cities Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 Luis Castiella International Workshop on Climate Change, Tourism and Education (CCTE) Sustainable Historic Urban Landscape: Valuation Tools . . . . . . . . . . . . . . . 355 Juan Camilo Rivera Dosman, Carlos Gabriel García Vázquez, Gaia Angelica Redaelli, and Julia Rey-Pérez Do Emotions Matter? Reviewing the Last Generation of Studies on Climate Change Communication and Tourist Behaviour . . . . . . . . . . . 367 Yen E. Lam-González, Carmelo J. León, Javier de León, and Mohamed Abderrahmane Ebnou The Training of Territorial Enhancers for the Valorization and Management of Biocultural Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378 Bernardo Rivera-Sánchez, Hellen-Charlot Cristancho-Garrido, and Andrés-Felipe Betancourth-López Rural Tourism Management as a Route to a Smart Territory Within the Framework of the Triple Transition. Opportunities from a Colombian Rural Territory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 Mónica Agudelo-López, Bernardo Rivera-Sánchez, Andrés-Felipe Betancourth-López, Sandra Olaya-Cardona, Felipe Aristizábal-Cardona, and Natalia Gallego-Gálvez International Workshop on Intelligent Systems Applied In Adaptive Smart Areas (ISAIAS) Monitoring System for Detecting Non-inclusive Situations in Smart Cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 Sebastian Lopez Florez, Alfonso González Briones, Juan Pavón, Rubén Fuentes-Fernández, and Juan Manuel Corchado

Contents

xxi

Evaluation of XAI Models for Interpretation of Deep Learning Techniques’ Results in Automated Plant Disease Diagnosis . . . . . . . . . . . . 417 Marco de Benito Fernández, Daniel López Martínez, Alfonso González-Briones, Pablo Chamoso, and Emilio S. Corchado Edge Service Allocation Based on Clustering Techniques . . . . . . . . . . . . . . 429 Marcelo Karanik, Iván Bernabé-Sánchez, and Alberto Fernández A Brief Review of Explainable Artificial Intelligence (XAI) Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442 Daniel López Martínez, Marco de Benito Fernández, Alfonso González-Briones, Pablo Chamoso, and Emilio S. Corchado Dynamic Transfer Point Allocation for Rural Demand-Responsive Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453 Pasqual Martí, Jaume Jordán, Fernando de la Prieta, and Vicente Julian Doctoral Consortium Convolutional Neural Network for DDoS Detection . . . . . . . . . . . . . . . . . . . 467 Fabian Ramirez, Gustavo Isaza, Néstor Duque, Jeferson Arango Lopez, and José Montes Trends and Applications of Heuristic Algorithms in Transportation Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474 Paola Alzate, Gustavo Isaza, Eliana Toro, and Jorge Jaramillo The Effects of Eco and Smart Policies: A Social Justice Perspective . . . . 480 Brian F. G. Fabrègue Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487

Main Track

Motorized Mobility on a Latin American University Campus: A Preliminary Study Focused on Sustainability Yamila S. Grassi1 , Mónica F. Díaz1,2 , Gabriela Pesce3 , Florencia Pedroni3,4 María Andrea Rivero3 , and Héctor G. Chiacchiarini5(B)

,

1 Planta Piloto de Ingeniería Química (PLAPIQUI, UNS-CONICET), Bahía Blanca, Argentina 2 Departamento de Ingeniería Química, Universidad Nacional del Sur (UNS), Bahía Blanca,

Argentina 3 Departamento de Ciencias de La Administración, Universidad Nacional del Sur (UNS) e

Instituto de Investigación en Ciencias de La Administración (IICA, UNS), Bahía Blanca, Argentina 4 Postdoctoral Fellow of the National Council of Scientific and Technical Research (CONICET), Bahía Blanca, Argentina 5 Departamento de Ingeniería Eléctrica y de Computadoras, UNS e Instituto de Investigaciones en Ingeniería Eléctrica “Alfredo Desages” (IIIE, UNS-CONICET), Bahía Blanca, Argentina [email protected]

Abstract. Universities are often viewed as microcosms of the cities they inhabit, making them ideal locations to study sustainable mobility solutions that could be implemented on a larger scale. The present work was performed at Universidad Nacional del Sur´s Altos de Palihue campus, in Bahía Blanca (Argentina). The research was based on an analysis performed on the community’s mobility preferences through video recordings at the campus access points and on a survey of the vehicle characteristics located at the parking lots. Based on this, a characterization of the university community’s way of commuting was obtained, as well as an estimate of the vehicle fleet’s age and the type of fuel used. According to the results, private combustion vehicles are dominant in both campus accesses, representing 45% using the access through the wooded neighborhood and 87% through the nearby avenue entrance. Interestingly, 64% of the motorized vehicles that access the San Andres entrance do so with only one occupant, while at the Cabrera entrance, this situation represents 67%. Moreover, based on the parking lots survey, 36% of the vehicles are less than 6 years old. Additionally, about 78% of the vehicles use petrol, while 18% are diesel and the remaining ones run on compressed natural gas. Last but not least, this study establishes the basis for a future emissions inventory generated by the current university community’s mobility decisions. In this sense, it will be possible to design and implement measures aimed at sustainable development, aligned with the Sustainable Development Goals. Keywords: Motorized mobility · University environment · Sustainable development · Sustainable development goals · Diagnosis

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 3–14, 2023. https://doi.org/10.1007/978-3-031-36957-5_1

4

Y. S. Grassi et al.

1 Introduction Currently, there is growing concern about urban pollution, mainly from mobile sources [1]. In addition, the daily traffic problems in cities are pushing the development of new trends in urban mobility [2]. In this sense, the shift in the way we move around the city implies a major cultural change. Today, a city that considers itself developed should no longer have streets crowded with motorized traffic. In this sense, sustainability-oriented societies promote active mobility and public passenger transport. In this way, urban transit places people as the main priority. This situation is aligned with the Sustainable Development Goals (SDGs) proposed by the United Nations, mainly associated with sustainable cities (SDG 11), climate change (SDG 13), health and well-being (SDG 3), and reduction of inequalities (SDG 10), among others [3]. The present study was carried out at the Altos de Palihue campus, belonging to the Universidad Nacional del Sur (UNS), in Bahía Blanca, Argentina. The city has 335190 inhabitants, according to the preliminary results of the last census in 2022 [4], and consequently, it is categorized as a medium-sized city. Likewise, it is considered a strategic intermediary location since it functions as a nexus between different regions of the country, through railways, roads, and airways, as well as with the rest of the world through its seaport [5]. Despite all this infrastructure, the city has urban buses as the only means of public transportation, with the private car being the dominant vehicle in the city’s downtown area, and within the local registered vehicle fleet [5–7]. The city is home to important universities, both public and private, and one of them is the UNS. Interestingly, the city adheres to the previously stated SDGs [8]. University campuses can be considered as a microcosm closely related to the cities in which they are placed since they are defined by the identity of the community that hosts them [9, 10]. University populations are usually composed of different socioeconomic backgrounds and ages, which generate irregular schedules and constant movement of people throughout the day [11]. In this sense, university campuses are deemed to be excellent study cases to evaluate sustainable mobility solutions that could then be implemented on a larger scale in its city [9, 11]. This situation makes it possible to analyze and evaluate the commuting patterns of the university community to design guidelines for more sustainable and intelligent mobility, which can be useful for decision-makers [12]. Although cities and universities have different objectives, they share a similar socioeconomic, environmental, and geographical context, so they have common infrastructures, transportation networks, and even similar challenges and needs [13]. Consequently, universities are expected to lead the sustainable mobility initiatives of cities [10]. In this context, this paper is framed in a research project, total duration of 18 months, that aims to multidimensionally analyze sustainable urban mobility proposals and collaborate with their implementation at a local level—Bahía Blanca, Argentina [2]. In particular, this work focuses on achieving a mobility diagnosis on the Altos de Palihue campus, belonging to the UNS. In this sense, this research raises the following questions: (1) What are the main means of transportation used by the UNS community? (2) What are the main characteristics of each type of motorized vehicle that arrives on campus? The findings obtained in answering these questions are useful to identify the most relevant factors to consider when designing and implementing a sustainable mobility framework for a university campus.

Motorized Mobility on a Latin American University Campus

5

2 Methodology The study of vehicle segmentation and flow provides essential information for generating policies and regulations to improve mobility, as well as promote ideas for change. In addition, a correct segmentation of the vehicle fleet allows for determining how the analyzed vehicle fleet is constituted, and different trends can be evaluated [14]. To fulfill the objectives of this paper, a relevant study of the amount and type of vehicles accessing the Altos de Palihue campus of the UNS is carried out. In the framework of a more extensive research project1 , this work aims to carry out a diagnosis of the current mobility of all the actors that arrive at the university campus. For this purpose, on-site videos were recorded at the two access points of the study area (see Fig. 1).

Fig. 1. Altos de Palihue Campus of the Universidad Nacional del Sur. The study area is delimited by red dashed lines, as well as the access points to the campus (both, at SA and at CA), which were the filming sites. Source Own elaboration

It should be taken into account that the characteristics of the campus accesses may affect the type of mobility used in each one. In Fig. 1, besides the entrance locations, the particularities of each access can be observed while, Fig. 2 shows the estimated distances between the different points of interest, such as the campus accesses, the Alem 1 The framework research project, total duration of 18 months, approved by the Inter-American

Organization for Higher Education, aims to multidimensionally analyze sustainable urban mobility proposals and collaborate with their implementation in Bahía Blanca (Argentina).

6

Y. S. Grassi et al.

campus of the UNS, and the Bahía Blanca city center. In this sense, the SA access is located in a quiet and wooded neighborhood, which favors walking. In addition, this entrance is closer to the UNS Alem campus, 2.7 km away by car (2.1 km by foot), which makes it easier to move between the two points, while the CA access is 3.6 km away by automobile (3.1 km walking). Furthermore, the SA entrance has a bicycle lane and bus stops, both public and free offered by the university between Alem and Altos de Palihue campuses. On the other hand, the CA access is linked to an urban highway such as Cabrera Avenue, and although it has bicycle lanes, this entrance is far from the university zone. In addition, the nearest bus stops are located five blocks away from the CA access. Finally, it can be said that the SA access is 3.5 km away from the city center by car (3.2 km walking) while the CA is 3.3 km distant by automobile (3.2 km on foot), and there is a distance of 1.4 km between the two accesses.

Fig. 2. Map of an area of Bahía Blanca city, where the following points of interest can be seen: city center, Alem campus, Altos de Palihue campus, and its accesses (SA and CA). Source Own elaboration

The videos were taken on September 6, 8, 12, 14, and 16, 2022, in five time periods of high vehicular flow (at 8:00, 12:00, 14:00, 16:00, and 18:00 h), which were previously explored through pilot studies on August 16, 17, and 22, 2023. The period under study is an intermediate month of the second academic semester, which means that educational activities are currently in full swing. In this sense, it is considered optimal to survey this month since it represents the actual commute of people attending the campus. The videos made at the San Andres entrance (see access point SA in Fig. 1) last approximately 35 min, while those generated at the Cabrera access (see access point

Motorized Mobility on a Latin American University Campus

7

CA in Fig. 1) have an extension of 25 min. This differentiation was made because, in the pilot studies, it could be observed that the time window in which people arrived through the CA access was smaller than through the SA one. In this sense, 30 h of recording were obtained. The videos were made using personal digital cameras. Taking into consideration the above information and the methodology used so far for obtaining the videos at both entrances to the Altos de Palihue campus, the study could be replicated in other months. However, it should be considered that the holiday months (January, February, and July) will not provide useful information for this kind of mobility research. Nevertheless, as was already noted, a month like the one chosen (September) presents the ideal conditions for conducting a mobility study. Once the filmed material has been captured, the different categories of mobility are counted manually through direct observation of the videos. In this sense, the different types of vehicles and people passing through the monitored points are counted and an attempt is made to establish the number of occupants of the different vehicles. The classification obtained is segmented into bicycles, electric bicycles, electric scooters, combustion motorcycles, electric motorcycles, private cars, private pick-up trucks, buses, cabs, light commercial vehicles, as well as people on foot, those who get on/off the bus or car/cab, whose vehicle does not enter the campus. It should be noted that this classification is carried out through a standardized protocol developed in a spreadsheet, which can be seen in Fig. 3. In addition, it is counted whether motorcycles come with one or two passengers and whether cars and vans arrive (or leave) on campus with one, two or more occupants. This allows us to know the number of people arriving by each available means of transportation. In addition, it provides an estimation of the number of vehicles traveling to or from the campus globally, considering the total population of the university community. It is important to note that using standardized information not only enables efficient data processing but also ensures that people performing the count do so as efficiently as possible. In this regard, the following protocols allow a different person to replicate this work and achieve the same results after counting by watching the videos. On the other hand, a survey of vehicle license plates was conducted in the different parking lots of the campus, to obtain an estimate on the age of the vehicle fleet that arrives at the campus. For this purpose, five inspections were carried out on November 18 and 25 and December 5, 6, and 12, 2022, between 14:00 and 17:00, in which the license plates of the vehicles parked on campus were categorized. For this, the methodology used in Grassi et al. [15] was followed, where it is considered that the type AA-000-AA and similar license plate format are assigned to the newest vehicles (since 2016), then the format from AAA-000 to PZZ-999 correspond to vehicles patented since 1995 while the domains from RAA-000 correspond to the oldest units (registered before 1995). Figure 4 shows the formats of the license plates used in Argentina. During these surveys, another segmentation of interest was carried out: the categorization of vehicles by fuel type used. In this sense, an attempt was made to identify whether the vehicle was a petrol, diesel, or compressed natural gas (CNG) one. The amount of electric and hybrid-electric vehicles in Bahia Blanca is still negligible, so they were not considered for the selected classification. The results obtained from these surveys are presented in a standardized form following the protocol shown in Fig. 5.

8

Y. S. Grassi et al.

Fig. 3. Standardized protocol for counting results presentation used in this paper. Source Own elaboration

All the data collected not only contribute to generating an initial diagnosis of mobility at the university campus, but also to analyze ideas for the development of more sustainable mobility. Likewise, in future work, such data will allow the estimation of the emissions generated by each type of vehicle in a more specific way, since it shows the average age of the vehicle fleet and the type of fuel they use.

Motorized Mobility on a Latin American University Campus

9

Fig. 4. Evolution of the Argentinean vehicle license plate from 1964 to the present. Source Own elaboration MOTORIZED VEHICLE SURVEY RESULTS PRESENTATION PROTOCOL Survey date Hour Number of cars surveyed Petrol

Diesel

CNG

Unidentifiable

Total

License plate type AA 000 AA License plate type AAA 000 (from AAA 000 to PZZ 999) License plate type AAA 000 (from RAA 000 to ZZZ 999) Total

Fig. 5. Standardized protocol for the presentation of survey results of motorized vehicles parked on campus according to their age and type of fuel used. Source Own elaboration

3 Results and Discussions The most relevant results of the mobility analysis carried out on the UNS Altos de Palihue campus are presented in this section. In this sense, the following is a subsection related to the segmentation by type of mobility used by the people who attend the campus. In addition, another section presents the results of vehicle fleet age on campus as well as the type of fuel used. 3.1 Global Characterization of University Campus Mobility The number of people entering and leaving the campus, segregated by transportation type, was obtained through manual counting by direct observation of the videos made at each entrance to the campus during the five daytime periods. In this sense, it was found that the San Andres (SA) access is the most used, reaching an average flow of 750 people during peak daytime hours, compared to 181 people at the Cabrera (CA) entrance. On the other hand, the mode of transportation most commonly used at each access point varied. Although in both entrances the private car, and in general combustion vehicles, are the dominant ones; at the SA entrance a large participation of public transport (24%), which includes the free bus service of the UNS between the Alem and the Altos de Palihue campuses, and active mobility (30%) (see Fig. 6), can be observed. These results are in line with a survey conducted among the university community where it was found that 40% of participants travel by private combustion vehicles, 32% on buses, 26% use active mobility, while only 2% go by electric mobility [2]. It is interesting to note that

10

Y. S. Grassi et al.

64% of the cars and pickup trucks that enter/leave through the SA access do so with only one occupant, while on CA this percentage is 67%. These results are interesting to know since they could be useful to generate alternative ideas on sustainable mobility, such as the implementation of an application that facilitates carpooling.

Fig. 6. Characterization of the means of transportation used by the UNS Altos de Palihue campus university community, by type of mobility, both for SA and CA access. Source Own elaboration

It is interesting to note that the characteristics of each entrance affect the type of mobility that circulates through them. In this sense, Fig. 6 shows that private combustion vehicles are the majority on access CA, which could be due to the proximity to Cabrera Avenue, a fast transit road for this type of vehicle. On the other hand, the mobility distribution in the SA access is consistent since it has a bicycle lane and a bus stop very close to the entrance (see Fig. 1), which encourages cycling (active mobility) and the use of buses, both public passenger transport and the UNS free service. In addition, walking is favored at the SA entrance due to the characteristics of the environment in which it is located, since the neighborhood to be crossed is quiet and wooded, which makes active mobility more pleasant. Moreover, the SA entrance is the closest to the UNS Alem campus (see Fig. 2), so those who need to travel between the two locations will use this option. Based on the above information, and considering that the population that attends the university campus both to study and to work is 18,507 people (15,412 undergraduate students), it can be estimated that a motorized vehicular amount of 7770 private vehicles drive to the campus with different frequencies and stay during the working hours (see Table 1). It should be remembered that in this study not only cars but also motorcycles, cabs, and pickup trucks (widely used in the city as private vehicles for commuting) are considered private combustion vehicles.

Motorized Mobility on a Latin American University Campus

11

Table 1. Estimation of the motorized vehicle fleet that accesses the university campus Participationa

Populationb

Percentage PCVc

People in PCV d Number of PCV e

SA entrance

80.56%

14909

45%

6709

5260

CA entrance

19.44%

3598

87%

3130

2510

9839

7770

Total

18507

a Population share entering through each access point to the campus b Population entering the campus by access point c Share of people traveling in private combustion vehicles (PCV) d Number of people using PCV per access point e Number of PCV per access point traveling to the campus, considering the number of occupants

per vehicle Source Own elaboration

3.2 Characterization of University Campus Motorized Vehicle Fleet On the other hand, through the research carried out in the campus parking lots, it was possible to obtain an estimate of the age of the motorized vehicle fleet that accesses the campus, as well as to characterize it by the type of fuel it uses. It is worth mentioning that motorcycles were not considered in this survey. In this sense, a segmentation of the motorized vehicle fleet is obtained according to whether it is more than 27 years old, up to 6 years old, or intermediate, and whether it is a petrol, diesel, or CNG vehicle. Table 2 summarizes the data obtained, considering that only 0.40% could not be classified by fuel type. From this analysis, it can be concluded that most of the motor vehicles that access the campus are petrol (~78%), followed by diesel (~18%), while CNG ones are in the minority (~4%). In terms of age, vehicles are less than 27 years old, representing 97% of the vehicles surveyed. It is interesting to note that a large percentage of vehicles are less than 6 years old (36%). Table 2. Segmentation of the motorized vehicle fleet that accesses the university campus, according to fuel used and estimated age. Petrol (%)

Diesel (%)

CNG (%)

Unclassified (%)

Total (%)

Up to 6 years old

28.24

7.09

0.40

0.00

35.73

Between 7 and 27 years

47.98

10.02

3.24

0.30

61.54

1.42

0.71

0.51

0.10

2.73

77.63

17.81

4.15

0.40

100.00

More than 27 years Total

Source Own elaboration Note Motorcycles are not considered

As mentioned above, all this obtained data will allow, in future works, to carry out a detailed emission inventory of campus mobility. This work is useful because campus

12

Y. S. Grassi et al.

mobility can be analyzed in a wide and diverse manner. In this way, decision-makers can implement new ideas to transform the university community’s mobility to move towards a more sustainable one. In this sense, it is well known that it has become a worldwide movement to encourage sustainable commuting to universities [16].

4 Conclusions Universities can be considered microcities that reflect what is happening at the city level in which they are located. Moreover, they are not only educational institutions but are also seen as leaders in terms of innovation, especially nowadays in aspects of sustainability. In this sense, it is known that university life generates a large number of trips and commuting for its community, which is why promoting more sustainable mobility is an objective to be achieved. This study analyzed the mobility used by the university community of the Altos de Palihue campus of the Universidad Nacional del Sur. The research was based on recording videos to obtain a characterization of how the university community moves around, as well as conducting surveys of the motorized vehicle fleet to estimate its age and the type of fuel used. The results obtained in this study reflect the specific conditions of the Universidad Nacional del Sur. However, the analysis of the mobility choices of the community provided by this study can help decision-makers to develop and improve policies and infrastructure that aim to encourage a more sustainable commuting mode to the university. According to the results, in general, private combustion vehicles are dominant at both entrances to the campus, representing 45% at the SA entrance and 87% at the CA entrance. On the other hand, based on the survey in the parking lots, it was determined that 97% of the motor vehicles are less than 27 years old and 36% represent vehicles less than 6 years old. In addition, it was determined that 78% of the vehicles surveyed use gasoline, while 18% are diesel and the remaining ones run on CNG. It is also interesting to note the percentage of motorized vehicles that access the campus with only one occupant, which is 64% for the SA entrance and 67% for the CA. In this sense, knowing that private motorized vehicles are predominant on campus measures, options such as carpooling could be implemented that encourage a more sustainable mode of private motor vehicle use. These actions could have the potential to move campus mobility to a more sustainable level. The strength of this study is to lay the groundwork for future work to quantify the air pollution generated by the current university community’s mobility decisions and evaluate improvements based on proposals for sustainable development, in particular promoting the SDGs on sustainable cities and communities (SDG 11), climate action (SDG 13), affordable and clean energy (SDG 7) and health and wellbeing (SDG 3). Acknowledgements. We are grateful to the Inter-American Organization for Higher Education (IOHE) for funding this research through the project entitled “Proposals for sustainable urban mobility from a multidimensional perspective” (code 029, item 2 in https://oui-iohe.org/es/ies-inv estigacion-colaborativa), and the Universidad Nacional del Sur for its financial support through the university extension project entitled “Sustainable urban micro-mobility from a multidimensional perspective for the city of Bahía Blanca”. Additionally, we extend our thanks to Universidad Nacional del Sur and CONICET for their support.

Motorized Mobility on a Latin American University Campus

13

References 1. Gulia, S., Nagendra, S., Khare, M., Khanna, I.: Urban air quality management—a review. Atmos. Pollut. Res. 6(2), 286–304 (2015). https://doi.org/10.5094/APR.2015.033 2. Pesce, G., Pedroni, F., Moral, P., Rivero, A., Grassi, Y., Chavez, E., Chiacchiarini, H., Sarro, L., Díaz, M., Finocchiaro, F., Alonso Revelli, F., Di Paolo, M., Saborido, T., Simone, A., López Hernández, A., López Hernández, G., Nascimento, L., Behr, A., González Martínez, J., Carrizo, F., Mayer, A., Frapiccini, R.: Movilidad urbana sostenible desde una perspectiva multidimensional. Caso: campus universitario de Palihue en Bahía Blanca. In: Pesce, G. (Ed.) Libro de resúmenes del 1° Encuentro de Investigación en Ciencias de la Administración. Instituto de Investigación en Ciencias de la Administración. Departamento de Ciencias de la Administración, Universidad Nacional del Sur, Bahía Blanca (2022). https://repositoriodigi tal.uns.edu.ar/handle/123456789/6273 3. United Nations—Sustainable Development Goals, https://www.un.org/sustainabledevelop ment/. Accessed 08 Feb 2023 4. Instituto Nacional de Estadística y Censos (INDEC) Preliminary results of the 2022 census. https://www.censo.gob.ar/. Accessed 06 Feb 2023 5. Grassi, Y., Brignole, N., Díaz, M.: Vehicular fleet characterisation and assessment of the on-road mobile source emission inventory of a Latin American intermediate city. Sci. Total Environ. 792, 148255 (2021). https://doi.org/10.1016/j.scitotenv.2021.148255 6. Grassi, Y., Brignole, N., Díaz, M.: Pandemic impact on air pollution and mobility in a Latin American medium-size city. Int. J. Environ. Stud. 79(4), 624–650 (2022). https://doi.org/10. 1080/00207233.2021.1941662 7. Grassi, Y., Díaz, M.: Post-pandemic urban mobility in a medium-sized Latin American city. Is sustainable micro-mobility gaining ground? Int. J. Environ. Stud. (2023). https://doi.org/ 10.1080/00207233.2023.2195327 8. Municipalidad de Bahía Blanca (MBB) Decreto 1273/2020—Programa. ODS BAHIA 2030. https://www.bahia.gob.ar/decretosyresoluciones/decreto/1/2020/1273/. Accessed 10 Feb 2023 9. Tomás, R., Fernandes, P., Macedo, J., Coelho, M.: Carpooling as an immediate strategy to post-lockdown mobility: a case study in university campuses. Sustainability 13, 5512 (2021). https://doi.org/10.3390/su13105512 10. Azzali, S., Sabour, E.: A framework for improving sustainable mobility in higher education campuses: the case study of Qatar university. Case Stud. Transp. Policy 6(4), 603–612 (2018). https://doi.org/10.1016/j.cstp.2018.07.010 11. Papantoniou, P., et al.: Developing a sustainable mobility action plan for university campuses. Transp. Res. Proc. 48, 1908–1917 (2020). https://doi.org/10.1016/j.trpro.2020.08.223 12. Ribeiro, P., Fonseca, F., Meireles, T.: Sustainable mobility patterns to university campuses: evaluation and constraints. Case Stud. Transp. Policy 8, 639–647 (2020). https://doi.org/10. 1016/j.cstp.2020.02.005 13. Torres-Sospedra, J., et al.: Enhancing integrated indoor/outdoor mobility in a smart campus. J. Geograph. Inf. Sci. 29(11), 1955–1968 (2015). https://doi.org/10.1080/13658816.2015.104 9541 14. Pérez, J., de Andrés, J., Borge, R., de la Paz, D., Lumbreras, J., Rodríguez, E.: Vehicle fleet characterization study in the city of Madrid and its application as a support tool in urban transport and air quality policy development. Transp. Policy 74, 114–126 (2019). https://doi. org/10.1016/j.tranpol.2018.12.002 15. Grassi, Y., Brignole, N., Díaz, M.: Hacia el desarrollo de una movilidad inteligente para la ciudad de Bahía Blanca: Primer enfoque sobre la caracterización de la flota vehicular del microcentro. In: X International Conference of Production Research, ICPR—Americas 2020,

14

Y. S. Grassi et al.

pp. 189–203. EdiUNS, Bahía Blanca (2020). https://www.matematica.uns.edu.ar/ipcra/pdf/ icpr_americas_2020_proceedings.pdf 16. Dehghanmongabadi, A., Ho¸skara, S.: Challenges of promoting sustainable mobility on university campuses: the case of Eastern Mediterranean university. Sustainability 10(12), 4842 (2018). https://doi.org/10.3390/su10124842

Smart Urban Governance Through Geoinformation: The Importance of Geoportals for City Interoperability Luiz Ugeda(B) and Isabel Celeste Fonseca JusGov, University of Minho, Braga, Portugal [email protected], [email protected]

Abstract. The article highlights the role of geoinformation in characterizing smart urban governance. To do so, it seeks to define how geoinformation fits into various areas of municipal management, such as environment, urban planning, transportation, agriculture, geology, among others. It enables the creation of maps, spatial data analysis, modeling of geographic phenomena, visualization of geographic information, decision making, and much more. The role of geoportals in this context is also highlighted, in order to visualize how these interactions, occur, as well as interoperability, characterized as a driving principle of smart urban governance as a system. The article also provides examples of global cities that efficiently use geoinformation and geoportals, always valuing data interoperability as a central element for its realization. Keywords: Interoperability · Geoportal · Geoinformation · Governance

1 Introduction The use of geoinformation in the intelligent governance of cities is becoming increasingly relevant, especially with the advent of geoportals. This type of platform offers governments and citizens access to geographic information and data about the city, allowing for better decision-making in relation to urban planning and resource management. Therefore, it is important to frame what defines smart urban governance and its relationship to geoinformation. The concept of geoportals and how these platforms work will be presented. We will highlight some global references of geoportals as public Geoweb, including the main characteristics of each city and how they use data through information sharing, user collaboration, and integration of different data sources through interoperability. Geoportals will be analyzed as a central tool for decision-making in relation to urban planning, resource management, and project monitoring, highlighting their main functionalities, challenges, and opportunities related to their use in urban management. The great challenge that arises is that in many cases, different government agencies and companies have their own geographic information systems, which can make it difficult to share information and collaborate. The interoperability of geoportals, within © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 15–24, 2023. https://doi.org/10.1007/978-3-031-36957-5_2

16

L. Ugeda and I. C. Fonseca

the concept of smart urban governance, is essential to ensure the integration of different data sources and allow for the creation of more comprehensive and integrated solutions for urban management.

2 The Concept of Smart Urban Governance and the Approach of Law Smart urban governance is an approach to urban management that relies on the use of technology and data to make cities more efficient, sustainable, and inclusive [7]. This approach seeks to integrate different areas of urban management, such as planning, transportation, security, environment, among others, using technological and information solutions. Smart urban governance involves the use of technologies such as the Internet of Things (IoT), geographic information systems (GIS), artificial intelligence (AI), Big Data, and others, to collect, analyze and share information about the city in real-time. This enables more informed and efficient decision-making regarding the use of resources, urban planning, and public services [15]. Smart urban governance also involves greater citizen participation in urban management using technological solutions that allow for collaboration, information sharing, and citizen engagement in decision-making [11]. In this way, smart urban governance aims to create a more connected, efficient, and inclusive city that meets the needs and demands of all its inhabitants. Through these technologies, it is possible to collect real-time data on different aspects of the city, such as traffic, energy use, air quality, among others. For example, sensors installed on streets and avenues of the city can collect data on traffic flow, allowing public managers to identify congestion points and take measures to improve vehicle flow. Similarly, sensors installed on lighting poles can collect data on electricity consumption, allowing public managers to identify areas where consumption can be reduced, and resources saved. Furthermore, smart urban governance also involves citizen participation in urban management [17]. To do this, technological solutions are used that allow for collaboration, information sharing, and citizen engagement in decision-making. For example, mobile applications can allow citizens to report issues such as potholes on streets, faulty street lighting, among others. This data can be integrated into urban management systems and used by public managers to take measures to solve the problems. In summary, smart urban governance works through the integration of different technologies and information systems to collect, analyze, and share relevant information about the city in real-time. This enables more informed and efficient decision-making regarding the use of resources, urban planning, and public services, creating a more efficient, sustainable, and inclusive city. There is no specific international standard for smart urban governance, which makes it difficult to understand its legal purpose. However, there are several international initiatives and standards that can be used as a reference for the implementation of smart urban governance. For example, ISO 37106:2018—Sustainable cities and communities—Guidance on establishing smart city operating models for sustainable communities provides guidelines for the implementation of sustainable smart city management

Smart Urban Governance Through Geoinformation: The Importance

17

models [8]. In addition, ISO 37120:2018—Sustainable cities and communities—Indicators for city services and quality of life establishes a set of indicators to assess the performance of cities in areas such as transportation, energy, environment, among others [9]. Other initiatives include the European Union’s Smart Cities Framework, an initiative created in 2011 that aims to support the development of smart city solutions across Europe. The framework provides a set of guidelines for the implementation of smart city solutions, including the use of advanced technologies, the integration of different areas of urban management, and citizen participation. It is based on six pillars: governance, people, economy, environment, mobility, and technology. Each pillar includes a series of measures and actions that can be implemented to develop smart city solutions in each area. The Smart Cities Council is an international non-profit organization that aims to support the implementation of smart city solutions worldwide. The council provides resources, tools, and guidelines to help cities develop and implement smart city solutions, including urban greenery management system [4] as well as promoting collaboration and knowledge exchange among cities.

3 The Use of Geoinformation by Cities Through Geoportals Geoinformation is a term that refers to geographic or spatial information, that is, information that is related to a particular geographical location on Earth. This information is usually collected and processed through technologies such as geographic information systems (GIS), remote sensing, GPS, among others [12]. It is used in various fields such as the environment, urban planning, transportation, agriculture, geology, among others. It allows the creation of maps, spatial data analysis [10], modeling of geographic phenomena, visualization of geographic information, decision-making, and much more. Geoinformation has become increasingly important in a world that is increasingly connected and has a larger amount of data available. Through it, it is possible to better understand the world we live in and create more efficient solutions for the challenges we face. It is a specific type of Big Data, as geotechnology gives data a locational characteristic, which brings a set of analysis possibilities and identifies ways to improve the quality of life for its inhabitants. Cities use geoinformation in various ways, in several areas of activity, as can be identified in the Table 1. These are just a few of the ways that cities can use geoinformation, allowing public managers to make more informed and effective decisions. In order to do so, this data needs to be consolidated in a way that assigns a logic to public management. In this context, Geoportals emerge, which are internet platforms that provide geospatial information from a specific area or region in digital format, allowing users to access and use this information interactively. This information can include maps, satellite images, geographic data, information on transportation, infrastructure, public services, among others. Municipalities have increasingly used Geoportals as a tool for smart urban management, as they allow for an integrated and interactive visualization of the city’s geospatial

18

L. Ugeda and I. C. Fonseca Table 1. Ways to implement geoinformation in Smart urban governance.

Axis of smart urban governance

Ways to implement geoinformation

Urban planning

Mapping and analyzing urban spaces, identifying the most suitable areas for new developments, assessing accessibility and mobility, predicting environmental impacts, among other activities [3]

Public services

Managing public services, such as waste collection, public lighting, urban cleaning, among others. Through geoinformation, it is possible to identify areas that need special attention and optimize the allocation of resources

Natural resources management

Managing natural resources, such as water and energy, and monitoring environmental changes, such as deforestation, erosion, and climate change

Public participation

Increasing public participation in urban decisions, allowing citizens to present their concerns and suggestions [2]

Data analysis

Analyzing large volumes of data collected by urban sensors, cameras, and other sources, providing valuable information for decision makers

Environment

Monitoring and managing the city’s natural resources, identifying risk areas, developing reforestation projects and conservation of green areas, evaluating air and water quality, among other activities [14]

Public safety

Monitoring and analyzing criminal activities, identifying the most critical areas, assessing the performance of public security teams, and developing more effective policing strategies

Transportation

Monitoring vehicle traffic, developing more efficient routes for public transportation, managing parking, and evaluating the effectiveness of infrastructure projects

data. It is possible to monitor and plan urban development, as well as provide useful information to the population, such as public transportation routes, points of interest, tourist information, among others. Therefore, Geoportals are essential tools for Smart Urban Governance, as they allow cities to efficiently manage their geospatial data and use it to make informed decisions, improve interdepartmental coordination, increase citizen participation [13], and plan emergency and disaster responses. It is important to note that municipalities that adopt Geoportals in their urban management should also consider issues of privacy and data security, as well as ensure accessibility and usability of the platform for all users, including people with disabilities and the elderly. Public managers need to be aware of the benefits and challenges of using Geoportals and work in partnership with society to maximize the potential of these tools in smart urban management.

Smart Urban Governance Through Geoinformation: The Importance

19

In Table 2, there are some global examples, in alphabetical order, of how Geoportals can be employed to exercise Smart Urban Governance. Table 2. Examples of Geoportals for the Exercise of Smart Urban Governance. City, Country

Ways to use the Geoportal

Barcelona, Spain

The geoportal of the city of Barcelona is called “Open Data BCN” (https://opendata-ajuntament.barcelona.cat/), which provides access to over 300 geospatial datasets about the city, including information on transportation, security, environment, and public services. The data is available in various formats, including CSV, KML, and GeoJSON To interoperate its data, the city uses an open data architecture and shares information among different government departments and agencies. Data collected by sensors and other sources is shared in an integrated data system, allowing departments to use each other’s data for informed decision-making. Additionally, the city collaborates with businesses and academic institutions to develop innovative technology solutions that meet the city’s needs This framework contributes to Barcelona being frequently cited as an example of Smart Urban Governance, due to its efforts to use digital technologies to improve the quality of life of its inhabitants and make the city more sustainable and efficient

Cape Town, South Africa The geoportal of Cape Town is the “City of Cape Town Open Data Portal” (https://web1.capetown.gov.za/web1/opendataportal/), which provides access to more than 200 geospatial data sets about the city, including information on transportation, environment, public safety, and public services. The data is available in various formats, including CSV, KML, and GeoJSON To interoperate its data, the city works closely with various government agencies and private sectors to collect, share, and use data for informed decision-making. The goal is to create a smarter and more connected city that meets the needs of its residents and adapts to changes in the urban environment This Geoportal structure contributes to making Cape Town an example of how a city can adopt technologies to enhance urban governance and improve the quality of life for its inhabitants (continued)

20

L. Ugeda and I. C. Fonseca Table 2. (continued)

City, Country

Ways to use the Geoportal

Copenhagen, Denmark

Copenhagen has a geoportal called Københavns Kommune (https:// www.opendata.dk/city-of-copenhagen), which is maintained by the Danish government and provides geospatial data for all of Denmark. In terms of interoperability, the city takes an open and collaborative approach, with a range of initiatives aimed at promoting collaboration between different government agencies and between the public and private sectors. For example, the city has established a Center of Excellence in Open Data to help promote collaboration around the use of open data. The city also participates in a range of national and international initiatives aimed at promoting interoperability and data sharing, such as the European Open Data Initiative. This Geoportal structure contributes to Copenhagen being frequently cited as one of the smartest and most sustainable cities in the world, with a range of ongoing Smart Urban Governance initiatives

Lisbon, Portugal

The geoportal of the city of Lisbon is “Lisboa Geodata” (https:// websig.cm-lisboa.pt/), which is managed by the Lisbon City Council and gathers various geospatial information, such as thematic maps, satellite images and geographic data. The portal provides information on infrastructure, cultural heritage, tourism, urban mobility, public services, and other topics relevant to urban management. The data from Lisboa Geodata is interoperable, following international geospatial data standards, allowing for integration with other systems and platforms, such as the city’s urban management systems. This framework structure contributes to Lisbon being defined as a city with Smart Urban Governance initiatives, as it has been adopting innovative technologies and solutions to improve the quality of life of its inhabitants, such as the implementation of traffic and air quality monitoring sensors, as well as energy efficiency and sustainability initiatives

New York, USA

The geoportal of New York City is “NYC Open Data” (https://ope ndata.cityofnewyork.us/), which provides access to over 2,300 geospatial datasets about the city, including information on transportation, safety, environment, and public services. The data is available in various formats, including CSV, KML, and GeoJSON. To interoperate their data, the city uses an open data architecture and information sharing between different departments and government agencies. Data collected by sensors and other sources are shared on an integrated data system, allowing departments to use each other’s data for informed decision-making. The city also collaborates with companies and academic institutions to develop innovative technology solutions that meet the needs of the city. This framework structure contributes to New York City being one of the leading cities in Smart Urban Governance, with a wide range of initiatives aimed at improving the quality of life of its residents (continued)

Smart Urban Governance Through Geoinformation: The Importance

21

Table 2. (continued) City, Country

Ways to use the Geoportal

Singapore

The geoportal of the city-state of Singapore is “OneMap” (https:// www.onemap.sg/), which provides access to over 50 geospatial datasets about the city, including information on transportation, housing, health, and public services. The data is available in various formats, including CSV, KML, and GeoJSON. To interoperate its data, the city uses a collaborative approach, working closely with various government agencies and private sectors to collect, share, and use data for informed decision-making. Additionally, the city employs a range of advanced technologies, including artificial intelligence, to analyze data and provide real-time insights to decision-makers. The goal is to create a smarter and more connected city that meets the needs of its inhabitants and adapts to changes in the urban environment. This Geoportal framework contributes to Singapore being globally recognized for its innovative approach to Smart Urban Governance

Sydney, Australia

The geoportal of Sydney is the City of Sydney’s Data Portal (https://data.cityofsydney.nsw.gov.au/), which provides a wide range of data and information about the city. The portal allows users to view and download data in different formats, including georeferencing files. The city has also adopted open standards for sharing data between different government systems and departments, which facilitates data interoperability. This framework contributes to Sydney being highlighted as one of the most innovative in terms of smart urban governance. The city has a strong commitment to sustainable development and improving the quality of life of its citizens

Tokyo, Japan

The Geoportal of Tokyo can be accessed through the national link https://www.gsi.go.jp/. The city has implemented an open data strategy, with the aim of providing access to various information and data, including geospatial. In addition, Tokyo has invested in data interoperability technologies to enable different systems and platforms to communicate and share information efficiently. This framework contributes to Tokyo being a reference in Smart Urban Governance, with several initiatives and strategies to improve the quality of life of its inhabitants through advanced technologies and intelligent data

22

L. Ugeda and I. C. Fonseca

4 Interoperability as a Principle for Smart Urban Governance Interoperability of geoinformation can be defined as a fundamental principle for Smart urban governance, as it enables different geospatial systems, databases, and applications to exchange information efficiently and effectively. This means that interoperability enables the connection between different databases, information systems, and technological platforms, allowing information to be shared among different departments and organizations involved in urban management. This is crucial to ensure an integrated view of different areas of the city, allowing decision-makers to have a broader and more accurate view of urban reality through the Geoportal. In addition, interoperability allows geospatial information to be combined with other sources of data within the Geoportal, such as IoT sensors, mobile devices, and social networks, further expanding the city’s capacity for real-time analysis and understanding. Interoperability of geoinformation is important for a municipality as it enables efficient management by integrating information from different systems and sources, which helps simplify management and decision-making, reducing the time and effort needed to access and combine data. It also improves planning by identifying trends, creating predictive models, and supporting strategic and long-term planning in city sectors such as transportation, environment, health, security, and education. By coordinating data from different sectors, communication and transparency can be improved, as data can be easily shared between different departments and between the government and the public. This helps increase public trust in municipal policies and practices. Interoperability, as a central concept for geoinformation operation within Geoportals, receives normative treatment for its standardization in various laws and technical standards. As examples of legislation, the Executive Order 12,906 (1994) [6] of the United States, which deals with Coordinating Geographic Data Acquisition and Access: The National Spatial Data Infrastructure, brings interoperability mechanisms to municipalities [1]. The same occurs with Directive 2007/2/EC of the European Parliament and of the Council of 14 March 2007 [5], which establishes a spatial information infrastructure in the European Community (INSPIRE) and guides European municipalities in this direction. We must also highlight ISO 19100, which deals with a set of technical standards that establish the foundations and requirements for geospatial data, including technical specifications for the acquisition, processing, storage, retrieval, dissemination, and use of geospatial information, as well as the initiative of the Open Geospatial Consortium (OGC), an international organization that develops and promotes open standards for the interoperability of geospatial technologies, allowing different systems to communicate and share geographic information and the OpenStreetMap (OSM), a collaborative mapping project that aims to create a free and high-quality world map. In addition to these international standards and practices, many countries have their own regulations related to the use of geoinformation by municipalities. It is important for public managers to be familiar with these regulations and apply them in their activities, ensuring the quality and effectiveness in urban management.

Smart Urban Governance Through Geoinformation: The Importance

23

5 Conclusions Cities are increasingly seeking smart solutions to improve the quality of life of their inhabitants. The use of geoinformation, its systematization through Geoportals, and data interoperability are essential for these solutions to be effective. The use of Geoportals for smart urban governance is a growing trend that is expected to expand in the coming years. Governments and companies are increasingly recognizing the importance of geoinformation for decision-making and for creating more efficient and integrated solutions for urban management. With the expansion of the use of public Geoweb, it is expected that cities will become more efficient, sustainable, and inclusive, providing a more pleasant and high-quality urban environment for their inhabitants. Geoinformation interoperability should be taken as a fundamental principle for the feasibility of Smart Urban Governance, as it allows for the integration of data between different systems and institutions, enabling more efficient decision-making and smarter city management. In addition, interoperability facilitates collaboration between different sectors and institutions, encouraging cooperation and information exchange. And this interoperable geoinformation must be equally understood in a principled way: the right to be informed geographically as a species of the genus "right to information". This legal dimension is important for deepening the importance of national spatial data infrastructure agencies as the basis of smart cities. These must be recognized as the basis of smart cities as a new utopia [16] and need to receive the necessary support to promote interoperability and the availability of quality geoinformation. This national-local dialogue, between agencies and municipal assignments, will allow cities to be managed in a more efficient, sustainable, and inclusive way, meeting the needs and expectations of citizens. In this scenario, lawyers may become important agents of this technological interoperability, deepened by the rapid advancement of Artificial Intelligence, which should bring, consequently, the possibility of obtaining centralized governance through Geoportals, enabling the construction of a more connected, efficient, and just city.

References 1. B˛akowska-Waldmann, E., Brudka, C., Jankowski, P.: Legal and organizational framework for the use of geoweb methods for public participation in spatial planning in poland: experiences, opinions and challenges. Quaestiones Geographicae. 37, 163–175 (2018). https://doi.org/10. 2478/quageo-2018-0032 2. Brown, G., Kyttä, M.: Key issues and research priorities for public participation GIS (PPGIS): a synthesis based on empirical research. Appl. Geogr. 46, 122–136 (2014). https://doi.org/10. 1016/j.apgeog.2013.11.004 3. Damieri, R.P., Rosenthal-Sabroux, C.: Smart cities. How to Create Public and Economic Value with High Technology in Urban Space. Springer, Heidelberg (2014) 4. Dawidowicz, A., Nowak, M., Gross, M.: Land administration system and geoportal service for the need of a fit-for-purpose national urban greenery management system (UGMS). The concept for the EU member state of Poland. Acta Sci. Pol. Adm. Locorum 21, 53–81 (2022). https://doi.org/10.31648/aspal.7454

24

L. Ugeda and I. C. Fonseca

5. Directive 2007/2/EC of the European Parliament and of the Council of 14 March: Establishing an Infrastructure for Spatial Information in the European Community (INSPIRE). Published in the Official Journal (2007) 6. Executive Order n. 12.906 of the White House: Coordinating Geographic Data Acquisition and Access: The National Spatial Data Infrastructure. Office of the Press Secretary, The White House, Washington, D.C (1994) 7. Goldsmith, S., Crawford, S.: The Responsive City: Engaging Communities Through DataSmart Governance. John Wiley & Sons (2014) 8. ISO: ISO 37106:2018-Sustainable Cities and Communities—Guidance on Establishing Smart City Operating Models for Sustainable Communities. International Organization for Standardization (2018) 9. ISO: ISO 37120:2018(en): Sustainable Cities and Communities—Indicators for City Services and Quality of Life. International Organization for Standardization (2018) 10. León, A., Felipe, L.: The political economy of spatial data infrastructures. Int. J. Cartogr. 4(2), 151–169 (2017). https://doi.org/10.1080/23729333.2017.1371475 11. Marvin, S., Luque-Ayala, A., McFarlane, C. (Eds.): Smart Urbanism: Utopian vision or false dawn? (1st Ed.). Routledge (2015). https://doi.org/10.4324/9781315730554 12. Ratti, C., Claudel, M.: The City of Tomorrow: Sensors, Networks, Hackers, and the Future of Urban Life. Yale University Press (2016) 13. Rzeszewski, M., Kotus, J.: Usability and usefulness of internet mapping platforms in participatory spatial planning. Appl. Geogr. 103, 56–69 (2019). https://doi.org/10.1016/j.apgeog. 2019.01.001 14. Stratigea, A., Kyriakides, E., Nicolaides, C.: Smart Cities in the Mediterranean: Coping with Sustainability Objectives in Small and Medium-sized Cities and Island Communities. Springer International Publishing (2017) 15. Tomás, M., Cegarra, B.: Actores y modelos de gobernanza en las Smart cities”, en URBS. Revista de Estudios Urbanos y Ciencias Sociales, vol. 6, núm. 2, pp. 47–62 (2014) 16. Townsend, A.M.: Smart Cities: Big Data, Civic Hackers, and the Quest for a New Utopia. WW Norton & Company (2013) 17. Zhang, S.: Public participation in the Geoweb era: Defining a typology for geo-participation in local governments. Cities 85, 38–50 (2019). ISSN 0264-2751. https://doi.org/10.1016/j.cit ies.2018.12.004

Digital Twin Application Methodology for the Improvement of Production and Service Systems. Application to Waste Management Processes Jhonathan Mauricio Vargas Barbosa(B) , Omar Danilo Castrillon Gomez, and Jaime Alberto Giraldo García Departamento de Ingeniería Industrial, Universidad Nacional de Colombia, Campus La Nubia Bloque Q Piso 2, Manizales, Colombia {jmvargasba,odcastrillong,jaiagiraldog}@unal.edu.co

Abstract. The urban domestic waste produced by population centers has generated a growing environmental problem in recent years. As a result, the waste management industry has significantly developed in the country. However, since these processes are highly manual, their efficiency is not enough to meet the demand.The previously described problems can be solved with an improvement in processes, and since Industry 4.0 and especially Digital Twins can be in charge of improving processes, it is necessary to develop a comprehensive methodology for generating these solutions that faces current and future challenges in the industry, especially in the waste management industry. This document proposes a definition of digital twins, structures the methodological development that includes an architecture and a layered framework, and a series of methodological steps for the application of the digital twin according to the level of scope defined by the end-user. It can be highlighted how digital twins allow greater efficiency and precision in waste management since they provide a virtual real-time representation of the waste management system and allow simulations and testing to optimize processes. Digital twins also enable more informed decision-making in waste management, as they provide a detailed visualization of the entire system, which facilitates the identification of problems and opportunities for improvement. Keywords: Digital Twins · Application methodology · Improvement of Production Systems and Services

1 Introduction The global concern for mitigating environmental impacts has generated the development of government policies that promote waste management processes, which help to reduce the polluting load generated by society. The above generates the need to design and optimize waste recovery processes, especially those that pertain to organic waste, which has been identified as a potential resource that can be converted into usable substances © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 25–36, 2023. https://doi.org/10.1007/978-3-031-36957-5_3

26

J. M. V. Barbosa et al.

through transformations mediated by microorganisms. In view of this, the present doctoral research project outlines the strategy for the development of a methodology for the improvement of production and service systems based on Digital Twins, that addresses current and future challenges of the waste management industry; specifically, the organic waste recovery processes. The utilization of solid waste is part of the development of strategies placed to generate a sustainable environment and society, as well as part of the millennium objectives. In Colombia, the use of the putrescible fraction of municipal solid waste (MSW) has increased in recent years. In fact, according to ECLAC et al. [1], around 12.9% of all generated solid waste is used in the country, the equivalent to approximately 3 kg per inhabitant per month, being composting the most used option for this. However, the effectiveness of these processes is affected by the incipient nature of the separation processes at the source. Delving into the efficiency of composting processes, authors such as Pineda et al. [2] carried out a series of experiences where they found that of all the organic matter disposed for composting, only 60% was transformed into compost, and the rest was emitted into the atmosphere in the form of water vapor and CO2 . The stated indicates why tools and methodologies to improve production systems and services such as Digital Twin should be sought. Digital Twin is an emerging technology that is based on a virtual dynamic model of a physical object or a system, that, through the use of sensors, captures information from the real system and responds to various stimuli. The concept of Digital Twin is composed of a physical system in real space, a digital system in virtual space, and the connections that link the virtual space with the real one [3]. Digital Twins can present a tangible value for companies with the creation of new revenue streams and by helping them answer key strategic questions. With new technology capabilities, flexibility, agility, and lower cost, companies can start the path to the creation of a Digital Twin with lower capital investment and shorter time-to-value than was ever available before. The question may not be whether to start applying Digital Twins, but where to start to get the most value in the shortest possible time, and how to stay ahead of the competition. What will be the first step and how to start the Digital Twin generation? [4]. There are a variety of research projects that highlight the manner in which Digital Twins plays an important role in supporting intelligent industrial systems, which can help drive a variety of decisions. However, considerable resources and time are required to create a Digital Twin of a particular process [5]. With the effective application of tools such as the Digital Twin, greater operating efficiency in composting processes is expected, as well as a decrease in processing times and in the amount of waste to be disposed of [6]. The application of Digital Twin in Latin America offers important opportunities for a sustainable environmental development [7]. Considering all of the reasons listed above, this document seeks to propose a Digital Twin application methodology that includes a definition proposal, an architecture and a framework of these items.

Digital Twin Application Methodology for the Improvement of Production

27

2 Methodology 2.1 Digital Twins, Definitions and Characteristics One of the most important tools from the industry 4.0 that is based on simulation is the Digital Twin, which is presented as a disruptive technology in the simulation and analysis of industrial processes, capable of extracting the benefits of digital plant transformation to the maximum. Digital Twins are “live” digital representations of devices and processes that make up a production or service process, connected to the real system that they represent through “Cyber-Physical Systems” (CPS). With the live information from the plant, the history of operations and maintenance, and the application of Machine Learning techniques, it is possible to obtain a high-precision model whose behavior closely resembles that of the real system [8]. Digital Twins create a virtual model of a physical entity, simulate the behavior of the physical entity by means of data, and have the characteristics of real-time synchronization, faithful mapping, and high fidelity through the means of real-virtual feedback, data fusion analysis, decision iteration selection optimization, to promote the interaction and integration of the physical world and the information world, and increase or expand new capabilities for the physical entity [9]. The term “Digital Twin” was coined in the 1970s, however the current use of the term dates back to a presentation by Michael Grieves in 2002 that discussed how a virtual model of a product could significantly assist in the management of the life-cycle of the product [9]. It has been 17 years since Dr. Grieves first proposed the terminology in 2003, and the last five years have seen the rapid development of the Digital Twin. However, up until now, there had been no formal or widely accepted definition, nor any unified creation and implementation process. Different industries and fields of application have different perspectives and methods. Through an in-depth review, it is found that the Digital Twin is gradually coming out of its infancy and entering a stage of rapid development where researchers begin to explore actual practices and technologies in the industry (Table 1). Table 1. Some definitions and reference models of the Digital Twin. Taken from [10]. Author and year

Definition

Grieves et al. [11] The Digital Twin is a set of virtual information constructs that fully describes a potential or actual physical product manufactured from the micro atomic level to the macro geometric level Tao et al. [12]

The Digital Twin combines the physical entity with the high-fidelity virtual counterpart and the two parts of the business with each other throughout the life cycle, it also integrates and converges data from multiple sources to generate a more accurate and complete information

Zhuang et al. [13] The Digital Twin refers to the process and method of describing and modeling the characteristics, behavior, formation process, and performance of physical entity objects using digital technology

28

J. M. V. Barbosa et al.

The original vision that set out to fully understand and reflect every aspect of the physical twin still has a long way to go. The fields of application of the Digital Twin are widely distributed, showing great vitality [14]. In fact, Jia et al. [10] carry out a bibliographical review where it is shown that a standard definition of Digital Twin has not yet been reached, but rather multiple definitions according to the general vision of the authors; A summary of these definitions is presented below. 2.2 Waste Management The environmental problem caused by inadequate waste management and deficient disposal of solid waste has acquired great importance within government policies in recent years [15]. Taking this into account, Integral Solid Waste Management Plans (ISWM) have been formulated and implemented by the municipal environmental authorities; This document contains many necessary policies, including the need to carry out the classification process of the waste generated by the community, to achieve an adequate disposal according to the type of waste and its use. Besides the stipulated, the identification of the mechanisms of use that are given to organic waste requires an adequate selection process that is not, necessarily, carried out by the cleaning companies that acquire the contract for the disposal of the waste generated in each of the cities. It is necessary, however, to implement strategies that tend to generate a culture of recycling and separation at the source, and that in turn, must be tied to a viable management program for different types of waste [16]. The consumption and production processes are directly related to the generation of solid waste [17]. Padilla and Trujillo [18] reveal that only 3.6% of Colombians perform recycling. Comparatively, this level of recycling at source is below that of a middleincome country such as Brazil, which reached 7.3% in 2006, and even below lowermiddle-income countries such as the Philippines, which in 2006 reached a percentage of 28% and Malaysia that reached a percentage of 9% in 2012; However, in recent years Malaysia has experienced a progressive growth in recycling and source separation, projecting a rate of 25% by 2020 Trujillo [18]. Population growth has generated an increase of waste generation in the country. The production of organic solid waste per inhabitant is around 0.7 kg/families per day [19]. Generating an environmental problem. In this regard, it is stated that the direct causes of the origin of the environmental problem of solid waste can be specified in: (i) the concentration of the population in cities; (ii) industrialization that favors the appearance of durable materials, and (iii) new consumption habits [20]. For this reason, the sustainable management of municipal waste is necessary in the impact phases such as planning, design, operation and closure of places and disposal sites. In Colombia, 11 million tons of solid waste are produced per year [21] and by weight, 65% of all the waste that’s generated corresponds to organic solid waste [22]. In the Latin American context, the coverage and quality of solid waste management is lower than the provision of other services, since in most countries there are no public policies and national plans for solid waste management. On average, in some cities in the region 70% of all the waste produced is collected (50–70% in small cities and 85–90% in large ones), which indicates that 30% of the waste produced is not collected. This could

Digital Twin Application Methodology for the Improvement of Production

29

represent between 20 and 25 million tons a year that end up in the streets, clandestine dumps and water sources. 2.3 Methodological Development Waste management processes in Colombia historically occur using recycling as a way of earning an income for those families who have no other way to get ahead, which denotes its highly artisanal nature where communities have major problems with its final disposal. There are industrial processes that can generate an added value to waste, such as the production of organic fertilizers. Technological tools are present, constantly changing the way in which some activities are carried out, the social conception of reality and the way of communication. Manufacturing is no stranger to technological advances, and it is for this reason that the industry 4.0 is being generated today, which is responsible for the new paradigms that are being established in manufacturing and service provision. The industry 4.0 has been considered a new industrial stage in which various emerging technologies are converging to provide digital solutions [23]. It is a global transformation through digital integration and intelligent engineering, and it is expressed as the next level of manufacturing in which machines will be redefined in the way they communicate and perform individual functions [24]. Industry 4.0 tools such as the Digital Twin offer the ability to gain deep insight into the internal operations of any system, the interaction between the different parts of the system, and the future behavior of its physical counterpart in a way that is actionable for their stakeholders. Users and interested parties [7]. This additional knowledge of the operation of the system, combined with the constant acquisition of data, makes production processes increasingly efficient, with constant feedback taking them to a new level of intelligent processes. After proposing a general panorama, some questions of special analysis are generated, such as: Is the implementation of the Digital Twin favorable to the improval of the organic waste management process? How do structural elements such as sensors, servers and other technological tools make the implementation of the Digital Twin difficult for small industries? Is it possible to implement a methodology that reduces the costs of applying Digital Twin prototypes? Up to what level of integration of the Digital Twin can be reached with the waste recovery processes considering that these have a highly manual nature? These and other questions that will gradually become evident in the investigative process, serve as a guide to direct this academic exploration and the achievement of the proposed objectives. Given the above, some variables are raised that may affect the applicability of the Digital Twin in production processes, some of these are affected by the social and technological context of the region. These relationships described are of special importance for the analysis and development of a Digital Twin application methodology. Finally, what is sought in this investigative process can be summarized as the generation of a methodological proposal for the application of Digital Twins that will help to improve the production and service systems, applied to the waste management industry and to evaluate this methodology proposed in areas of high population density as follows:

30

J. M. V. Barbosa et al.

As consequence of the stated, the following general research hypothesis was proposed: With the development and application of an integral methodology for the adoption of Digital Twins in the processes of use of organic waste in Colombia, it is possible to improve the efficiency of said processes, measurable with different KPI’s. In the results section, a proposal for the definition of Digital Twins is exposed and based on the particularities of these for the processes of improvement of production systems and services, specifically for those in charge of the use of organic waste, an architecture is also developed, a framework and a methodology that facilitates its application.

3 Results 3.1 Digital Twin Definition Despite its conception since the 1970s, a formal or widely accepted definition has not been achieved, and neither has a unified creation and implementation process. Different industries and fields of application have different definitions, perspectives, and methods according to their application frameworks [14]. On account of this, it is pertinent to formulate a broad and structured definition of Digital Twins, which encompasses the characteristics of the application context, methodological generalities, and that generates a broad vision of its framework. A Digital Twin is a technique for the active representation of a real system with constant feedback, which, using autonomous learning tools, Data Mining and sensors, among other tools integrated into industry 4.0; generates active and predictive information from systems in virtual space. Depending on the level of implementation and the objectives set in the creation, the Digital Twin can be used as a tool for decision-making in the short, medium and long term, as an evaluation and training tool or a key tool for improvement of production systems. A proposed schematic definition of Digital Twins is shown in Fig. 1. The development of a Digital Twin must be planned following some key concepts and practices of project management. As in the simulation of discrete events, the Digital Twin should not be applied indiscriminately, but it is necessary to decide based on some considerations about the study system and the relevance of the problem to be solved with this tool. Given the above, it can be said that every real system can have its partner in the virtual world, mediated under the unidirectional transfer of information (data) in its primary state or bidirectional in its most developed state; where the virtual pair or Digital Twin has some characteristics and services generated based on the level of implementation where it is possible to go from a basic virtual copy to an intelligent decision-making system mediated by the constant generation and acquisition of data. The level of implementation depends directly on the objectives set out in the methodological implementation of Digital Twin. At a more advanced level of development of a DT, it can be used to make decisions and to modify the real system, therefore generating a feedback loop that is predictive and constant, as well as improving the studied systems; All of which is dependent on

Digital Twin Application Methodology for the Improvement of Production

31

Fig. 1. Scheme for proposed definition for Digital Twins

the level of implementation and the objectives of the Digital Twin itself that is to be developed. 3.2 Architecture and Framework Given the large amount of information that can be generated in the interaction between the physical twin and the virtual twin, it is necessary to create a series of mechanisms that facilitate information management. The architecture encompasses a wide variety of tools, processes, models and mechanisms to carry out its task in data management. The proposed architecture is composed of three fundamental layers, one in charge of the physical environment, another of the virtual environment, and a third layer of communication and data transfer. Both the physical environment layer and the virtual environment layer have structural and functional elements that describe the process, capture data, and act on the data processed and captured by the communication layer (Fig. 2).

Fig. 2. Proposed architecture.

32

J. M. V. Barbosa et al.

The proposed Framework is based on the architecture described above and it is made up of three fundamental elements: The physical environment, the virtual environment, and the online and offline services responsible for information processing. In the first place, the data of the real system is collected with the help of sensors and manual feeding, then this information is stored and processed by different logical-mathematical tools to later be modeled and used in the virtual system where it can feed back into the real system and help to optimize it looking for a better performance of it (Fig. 3).

Fig. 3. Digital Twin framework proposed for the improvement of production and service systems.

3.3 Methodological Proposal The proposed methodology is made up of a series of essential phases for the development of Digital Twins, starting at the relevance of this tool, and following through the structured definition of the objectives and services sought, until a functional Digital Twin that meets the objectives and services that were established in the project is reached, as evidenced in Fig. 4. The proposed methodology is structurally made up of two stages, the contextual stage and the purposeful stage. In the contextual stage, the aim is to formulate the Digital Twins application and implementation project. For this stage, as a first step, it is necessary to define the relevance of the Digital Twin for the application case. Secondly, it is necessary to clearly formulate the problem since the objectives and the level of application and implementation are based on this information in the proposal stage as well as the fact that the application project is formulated taking into account the particularity of each case. The proactive stage is limited by the level of application, and this depends on the objectives set in the contextual stage. For this methodology, three levels of application

Digital Twin Application Methodology for the Improvement of Production

33

Fig. 4. Methodological proposal developed for the implementation of Digital Twins.

are proposed: the first and the simplest is the digital copy of the real system. The second, which increases its level of complexity, implies a constant supply of data from the real system, and the third level of application has a particularity, being as the intervention of the real system from the virtual system is proposed, closing a circuit of information action. This stage begins in the project formulation phase, then goes through a phase of evaluation and classification of the studied system, continuing with the project, a phase of construction of the virtual system is proposed and a functional model of it is integrated into this virtual system that represents its real behavior. When the virtual and

34

J. M. V. Barbosa et al.

functional system are integrated, the interconnections between the real system and the virtual system are designed and thus are able to propose self-management and automation strategies that close the feedback loop allowing the analysis of the information generated regarding the operation, the functioning and the improvement of the studied system. The results presented in this document are focused on systems for organic waste management, although its generalities and methodology have the capacity to be applied to any production process susceptible to improvement and optimization.

4 Future Work For future work, we suggest comparing the proposed methodology with similar ones, integrating machine learning and/or artificial intelligence techniques, evaluating the scalability of the methodology, and exploring the application of digital twins in critical decision-making situations in real-time.

5 Conclusion Once this study has been completed, the design of a digital twin application methodology for the improvement of production systems represents a fundamental tool for the optimization of production processes. Through the use of digital twins, it is possible to simulate production processes virtually, which allows the identification of possible problems and makes it possible to evaluate solutions before their implementation in reality. Done this way, costs can be reduced, product quality can be improved and production efficiency increased. It is important to highlight that the implementation of this methodology requires a multidisciplinary team highly specialized in the subject, as well as an adequate investment in technology and resources. However, the benefits that can be obtained are significant and make it possible to improve the competitiveness and sustainability of the company. In short, the application of digital twins in the improvement of production systems is a growing trend that represents a key tool for the digital transformation of the industry. Currently, it is necessary within the generated methodology, to propose a definition that brings Digital Twins closer to the application, an architecture based on layers, in the case of this document, based on the 3 layers proposed: the first where the physical twin is housed, the second where the digital twin is housed and the third, an interconnection layer that facilitates the transfer of information. Based on the proposed architecture, a framework has been designed following the same layering strategy and providing a common and coherent structure for the design and implementation of the Digital Twin. Additionally, in the present work, a Digital Twins application methodology has been developed for the improvement of production and service systems, which consists of two stages in order to facilitate the application of Digital Twins for the improvement of production processes. This application depends directly on the application levels. Acknowledgements. The authors would like to express their sincere gratitude to Universidad Nacional de Colombia campus Manizales, for their support and resources in conducting this research. We also wish to thank the Faculty of Engineering and Architecture for their support

Digital Twin Application Methodology for the Improvement of Production

35

of the doctoral program in Industrial Engineering and Organizations. This article is part of the doctoral thesis entitled "Methodological Proposal for Improving Production and Service Systems in the Waste Industry through the Use of a Digital Twin. Application in High-Population Density Areas", developed within the framework of the doctoral program. We are grateful to the Ministry of Science, Technology, and Innovation for their financial support of this research, and to all participants and collaborators who made this study possible.

References 1. Comisión Económica para América Latina y el Caribe (CEPAL), Departamento Nacional de Planeación (DNP), Compromiso Empresarial para el Reciclaje (CEMPRE) Colombia, «Encuesta A Municipios Sobre Gestión De Residuos Sólidos Domiciliarios 2019 Colombia,» CEPAL, Bogotá (2021) 2. Vargas Pineda, O.I., Trujillo González, J.M., Torres Mora, M.A.: El compostaje, una alternativa para el aprovechamiento de residuos orgánicos en las centrales de abastecimiento. Orinoquia 23(2) (2019) 3. Grieves, M.: Digital Twin: Manufacturing Excellence through Virtual Factory. White paper Michael W. Grieves, LLC. Dassault Systeme, pp. 1–7 (2014) 4. Parrott, A., Warshaw, L.: Industry 4.0 and the digital twin, Manufacturing meets its match. About Deloitte University Press (2017) 5. Gao, Y., Chang, D., Chen, C.-H., Xu, Z.: Design of digital twin applications in automated storage yard scheduling. Adv. Eng. Informatics (2022) 6. Marmolejo, L.F., Oviedo, É.R., Jaimes, J.C., Torres, P.: Influencia de la separación en la fuente sobre el compostaje de residuos sólidos municipales. Agronomía Colombiana 28(2), 319–328 (2010) 7. Botín-Sanabria, D., Mihaita, A.-S., Peimbert-García, R., Ramírez-Moreno, M., RamírezMendoza, R., Lozoya-Santos, J.: Digital twin technology challenges and applications: a comprehensive review. Remote Sens. (2022) 8. Instituto Tecnológico de Informática: Digital twins: gemelos digitales en la transición a la Industria 4.0. 20 Octubre 2018. https://www.iti.es/proyectosidi/proyecto-gemelos-digitalesindustria-4-0/ 9. Li, L., Lei, B., Mao, C.: Digital twin in smart manufacturing. J. Ind. Inf. Integr. (2022) 10. Jia, W., Wang, W., Zhang, Z.: From simple digital twin to complex digital twin Part I: a novel modeling method for multi-scale and multi-scenario digital twin. Adv. Eng. Informatics (2022) 11. Grieves, M., Vickers, J.: Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems. de Transdisciplinary Perspectives on Complex Systems, Berna, Springer, pp. 85–113 (2017) 12. Tao, F., Zhang, M.: Digital Twin Shop-Floor: A New Shop-Floor Paradigm Towards Smart Manufacturing, pp. 20418–20427. IEEE (2017) 13. Zhuang, C., Liu, J., Xiong, H., Ding, X., Liu, S., Weng, G.: Connection, architecture and trends of product digital twin. Comput. Integr. Manuf. Syst. 23(4), 753–768 (2017) 14. Liu, M., Fang, S., Dong, H., Xu, C.: Review of digital twin about concepts, technologies, and industrial applications. J. Manuf. Syst. 346–361 (2021) 15. Avendaño Acosta, E.F.: Panorama actual de la situación mundial, nacional y distrital de los residuos sólidos. Análisis del caso Bogotá D.C. Programa basura cero, Bogotá D.C. (2015) 16. Rodríguez Gutiérrez, L.: La generación de residuos orgánicos en Cundinamarca y sus mecanismos de aprovechamiento en la generación de energías limpias. Universidad Cooperativa de Colombia (2020)

36

J. M. V. Barbosa et al.

17. Pandey, R.S.A.K.M. Exploring linkages between sustainable. J. Clean. Prod. 49–59 (2017) 18. Padilla, A.J., Trujillo, J.C.:: Waste disposal and households’ heterogeneity. Identifying factors shaping attitudes towards source-separated recycling in Bogotá, Colombia. Waste Manage. 74, 16–33 (2018) 19. Castañeda-Torres, S., Rodríguez-Miranda, J.P.: Modelo de aprovechamiento sustentable de residuos sólidos orgánicos en Cundinamarca, Colombia. Universidad y Salud, pp. 116–125 (2017) 20. Montes Cortés, C.: Estudio de los residuos sólidos en Colombia. Universidad Externado de Colombia, Bogotà D.C. (2018) 21. WWF: Por qué seguimos sin reciclar en Colombia? https://www.wwf.org.co/?363591/Porque-seguimos-sin-reciclar-en-Colombia 22. DANE: Cuenta ambiental y económica de flujos de materiales—residuos sólidos 2017–2018. Bogotá (2020) 23. Frank, A.G., Santos Dalenogar, L., Ayala, N.F.: Industry 4.0 technologies: implementation patterns in manufacturing companies. Int. J. Prod. Econ. 15–26 (2019) 24. Muhuri, P.K., Shukla, A.K., Abraham, A.: Industry 4.0: a bibliometric analysis and detailed overview. Eng. Appl. Artif. Intel. 218–235 (2019)

Adaptive and Intelligent Edge Computing Based Building Energy Management System Sergio M´ arquez-S´ anchez1,2(B) , Sergio Alonso-Roll´ an1,2 , Francisco Pinto-Santos1 , Aiman Erbad3 , Muhammad Hanan Abdul Ibrar3 , Javier Hernandez Fernandez4 , Mahdi Houchati4 , and Juan Manuel Corchado1,2 1

BISITE Research Group, University of Salamanca, Calle Espejo s/n. Edificio Multiusos I+D+i, 37007 Salamanca, Spain {smarquez,sergio.alro,franpintosantos,corchado}@usal.es 2 Air Institute, IoT Digital Innovation Hub (Spain), 37188 Salamanca, Spain 3 Information & Computing Technology Division. College of Science and Engineering, Hamad Bin Khalifa University, A 036-F LAS, Ar-Rayyan, Qatar {AErbad,mibrar}@hbku.edu.qa 4 Iberdrola Innovation Middle East, Doha, Qatar https://www.bisite.usal.es/es

Abstract. Most building and energy management system (BEMS) solutions follow a set of rules (supervised or unsupervised learning) to make energy-saving recommendations to inhabitants. However, these systems are normally solely trained on energy data meaning that they do not consider other key factors, such as the inhabitants’ comfort or preferences. The lack of adaption to inhabitants renders these energy-saving solutions largely ineffective. Moreover, BEMS solutions are cloud-based entailing greater cyberattack risks and a high data transmission load. To address these problems, this research proposes an edge computing architecture based on virtual organizations and distributed explainable artificial intelligence (XAI) algorithms for optimized energy use in buildings/homes and demand response. Thanks to virtual organizations’ energy efficiency (EE) measures, which consider the inhabitants’ comfort and dynamically learn from real-time inhabitant data, the consumption patterns of the inhabitants are effectively optimized. Keywords: Edge computing · Virtual organizations computing · Deep learning · Explainable AI

1

· Social

Introduction and Motivation

The demand for commercial and suburban settlement grows at a much higher pace than that of the efforts to increase the sustainability of buildings. It is QNRF—National Priority Research Program. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 37–48, 2023. https://doi.org/10.1007/978-3-031-36957-5_4

38

S. M´ arquez-S´ anchez et al.

estimated that in 2018, 20% of total global energy consumption came from buildings, specifically suburban and commercial properties. Moreover, it has been found that these buildings accounted for 28% of energy-related CO2 emissions on global scale. These statistics evidence a critical need for low-carbon, energyefficient buildings becoming a standard in towns and cities across the globe. This would also substantially contribute to preventing further global warming, as outlined in sustainable development goal 13, which seeks to limit temperature rise to below 2 ◦ C by 2030. The concept of smart homes aims to not only create environmentally-friendly buildings but also promote savings for both inhabitants and distribution network operators (DNOs). Smart buildings are enabled by technologies within the paradigms of the internet of things (IoT) technologies, as well as machine learning (ML) and deep learning (DL). IoT-based energy management systems (EMS) combine sensors, communication protocols and advanced learning algorithms for the collection of energy and data from homes or industries enabling information-driven efficient energy use in buildings/homes. There have been numerous ML and DL proposals for energy efficiency in buildings in recent literature. These measures are designed to optimize demand response (DR) through energy consumption and generation forecasts (including distributed energy resources—DER) under different operational circumstances. However, their internal analyses cannot be interpreted; they have a black-box system that identifies the relationship between various inputs and outputs by means of a number of supervised and unsupervised learning algorithms. However, these approaches usually have two major drawbacks: – The energy efficiency (EE) measures recommended by these ML and DL models solely consider energy-related variables. They are not focused on the inhabitants’ ability to adapt those solutions in their daily life i.e. they do not consider the inhabitants’ comfort or the reputation of the recommended energy-saving measures among inhabitants. – Generally, classical ML models, especially those based on DL, cannot be interpreted, impeding the recommendation of EE measures. Further advancements in smart buildings have been enabled by the emergence of cloud computing which makes it possible to control connected devices remotely. Unfortunately, the increasing number of IoT devices, network bandwidth and security weaknesses have become a bottleneck, limiting the performance of these systems. Edge computing is able to support cloud computing solutions, solving the problem of scalability and increasing cybersecurity. It is a new, promising computing paradigm that involves the creation of a network infrastructure in the user’s vicinity, which is plentiful in IoT resources (i.e. storage, compute, and bandwidth). Moreover, it minimizes latency by processing real-time, security-crucial data in a one-hop manner. Greater security is provided to sensitive data by processing it at the edge of the network, further security can be added by integrating distributed ledger technologies (DLT). In this research, the social computing approach is adapted to provide solutions to the problems discussed above. An architecture based on virtual organizations has been built to propose EE recommendations which encompass the

Adaptive and Intelligent Edge Computing Based Building

39

inhabitants’ preferences. The system makes these suggestions to users on the basis of analyzing similar consumption behavior profiles. Moreover, the blackbox problem is addressed through the integration of explainable artificial intelligence (XAI) approaches, such as hybrid neuro-symbolic AI algorithms, which increase interpretability, specifically, deep symbolic learning. This research article is organized as follows: Sect. 2 presents an overview of the state of the art in building and energy management system (BEMS). Section 3 presents the architecture of the system. Section 4 presents the conducted case study and some preliminary results. Finally, in the last section, conclusions are drawn and future lines of research are discussed.

2

State of Art

Smart home and smart building technologies are a current trend in the development of modern society that provides intelligent living environments for daily convenience and comfort. A smart building can be defined as a building containing a communications network, linking sensors, smart domestic appliances, and devices, that can be remotely monitored and controlled to provide services that respond to the behaviour of residents. The main objective of a smart building is to allow people to live in a convenient (maximize the residents’ comfort), economic, healthy and environmentally friendly manner. This goal may be achieved by deploying fully automated control of appliances in order to produce high levels of comfort and security, facilitate energy management, reduce environmental emissions and optimize energy consumption. Smart buildings have five essential characteristics: automation (ability to perform automatic functions), multi-functionality (ability to perform various duties), adaptability (ability to learn and predict residents’ behaviour), interactivity (ability to interact with different stakeholders such as building owner, network operator, etc.) and efficiency (ability to save time, energy and costs) [1,2]. Another closely related concept is that of a smart city. A smart city would be made up, for the most part, of smart buildings such as those described above. Current cities are experiencing constant growth, as indicated in [3]. Since 2007 the urban population has been exceeding the rural population and according to forecasts, it is expected that by 2050, 70% of the population will live in urban environments. This growth means that urban management is becoming increasingly complex. This fact, together with the need to optimise natural and energy resources, has led to the exploration of new, technological buildings, giving rise to numerous studies, pilot tests and even projects in production. Among notable studies in the field of smart buildings, [4] aimed, through the creation of IoT nodes integrated in a distributed environment, to generate real-time data to monitor the activity of an inhabitant within their home in order to predict the well-being of the individual. Another relevant study [5] proposed the use of techniques for monitoring energy consumption and tracking electricity market movements in order to carry out residential energy management where the home itself acts as a price-taking agent in the local market. Another example is GreenVMAS, presented in [6]. This system used the residual energy generated by

40

S. M´ arquez-S´ anchez et al.

power plants to heat greenhouses. There are also studies dedicated to recognizing the main parameters that should be monitored in the energy management of buildings, as is the case of [7], which explored the energy consumption patterns associated with human occupation using Internet of Things techniques. Determining these parameters and standardizing them can be decisive for the success of a project in this field. Using this knowledge as a basis, it is possible to build more robust projects with a more reliable scientific basis. A different approach was proposed in [8], where the aim was not to use IoT and sensorisation techniques but to improve energy efficiency directly by changing the behavior of users through education. To this end, the authors’ studied people’s behavior, warning them and correcting the behaviors that caused higher energy consumption. Leaving aside the numerous studies found in the field of smart buildings, we have carried out an analysis of the most promising technologies or disciplines for this sector, among which is the internet of things, better known as IoT. As described in [9], IoT is an emerging discipline in fields such as smart cities, smart homes, physical security, e-health, logistics etc. Architectures that enable the internet of things to be integrated into smart buildings have been discussed in [9]. In IoT there are different variants of distributed architecture, some studies, such as the one mentioned above or the study in [10], define a generic and distributed architecture of the internet of things, applied to the concept of smart cities and smart buildings. The internet of things (IoT) generates large amounts of data in all areas where it is applied, allowing smart buildings to become sources of data. Due to the variety and number of different IoT devices that can be applied, there are studies, such as the one presented in [11], dedicated to the transformation of heterogeneous data, collected by an IoT network in a smart building, into a homogeneous dataset. Linked to the term internet of things is a newer term known as the internet of energy or IoE. In [12], this concept was defined as a concrete implementation of the IoT, applied to distributed energy systems that seek energy efficiency and aim to protect the environment and avoid unnecessary energy waste. In the field of artificial intelligence, there are different studies and developments that are beginning to apply this type of techniques to energy saving. Among these studies is [13], which compiled works from numerous reliable sources such as Scielo, Dialnet or Elsevier to determine the different applications of artificial intelligence in energy saving, defending the viability of these techniques. We have also found reviews such as the one in [14], which showed the importance of energy saving and the need to use intelligent models aimed at reducing electricity consumption. Some examples of these applications are found in [15], where the use of a predictive control model in an energy management system is proposed, developing a comfort temperature predictor for HVAC systems. The development of smart buildings requires the deployment of disruptive technologies among which are big data engineering and analytics, artificial intelligence and machine learning, cloud and edge Computing, distributed sensing and actuation technologies such as the IoT, etc.

Adaptive and Intelligent Edge Computing Based Building

3

41

Architecture

This section presents the architectural proposal for the system. It starts with the division of architectural layers and the types of nodes deployed in them. Subsequently, these are broken down into the components to be developed and a technological proposal for their implementation is presented. 3.1

Layer architecture

As part of the architecture, three layers have been defined, with their respective nodes, each of which houses certain responsibilities in the system. For more detail, the layers are represented in Fig. 1. However, each node and responsibilities are enumerated as follows:

Fig. 1. Architectural proposal of layers, nodes, and deployment.

– IoT layer: The main characteristic of the components located in this layer is that they are deployed in the field. This layer is responsible for ingestion and communication at a low level, obtaining data from the different sources of information in the environment, mainly sensors. These make up the physical intake devices, which also belong to this layer. It is worth noting that the nodes in this layer are called “IoT nodes”, and that they are physical devices. Specifically, physical intake devices include sensors and sensor aggregation probes. Also in this category are the devices necessary for the communication of the IoT nodes with the edge node, such as a router, gateway, and other network device.

42

S. M´ arquez-S´ anchez et al.

– Edge layer: In this layer is the data management system for the data ingested from the IoT layer and it is responsible for preprocessing the data and applying artificial intelligence analysis to them. This makes it possible to later send the data to the cloud layer to ensure data persistence. The edge layer is housed in one or several computing nodes, deployed in several buildings’ rooms, close to the physical devices for data ingestion (probes with a set of integrated sensors). It is important to remark that the nodes described within this layer are called “edge nodes”, and in this category are the nodes deployed in the field, in charge of centralizing the data intake. They correspond to the edge layer, and contain the functionality needed to ingest the data sent by the IoT nodes, preprocess it, analyze it (using federated learning) and send it for storage in the cloud nodes. – Cloud layer: This layer houses the functionalities related to data persistence, coordination of artificial intelligence analysis and management of content that can be viewed by the user. As its name indicates, it is in a cloud environment to be determined. It is worth noting that the nodes described within this layer, are called “cloud nodes”. In this category are the nodes deployed in the cloud to perform the persistence, coordination of analysis and visualization of IoT data sent by the Edge nodes. 3.2

Multi-agent Architecture

For a more detailed definition, the architecture has been defined in the field of multi-agent systems, as can be seen in Fig. 2. The list of agents that make up this system can be traced to the services presented in previous sections, however, they are outlined below:

Fig. 2. Definition of architecture through the paradigm of multi-agent systems

– Virtual device organization: it can be composed of several replicated virtual organizations. Represented in blue color, it can be formed by the following agents:

Adaptive and Intelligent Edge Computing Based Building

43

• Edge ingest agent: agent in physical devices that is in charge of taking data and sending it to its associated virtual edge organization. – Virtual edge organization: it can be composed of several replicated virtual organizations. Represented in green color, it can be made up of the following agents: • Edge analysis agent: performs the simple analyses with little computational load on its virtual edge organization. • Edge coordination agent: coordinates communications and analysis in the virtual edge organization. It also coordinates data delivery to the virtual cloud organization. – Virtual cloud organization: represented in orange color, conformed by the following agents: • Communication coordination agent: coordinates the communications in the virtual cloud organization. • Cloud ingest agent: coordinates the preprocessing of the data ingestion coming from the virtual edge organization, then sends the preprocessed data to the cloud persistence agent. • Cloud persistence agent: in charge of storage management in the virtual cloud organization. • Cloud analysis agent: in charge of the management of the analyses performed in the virtual cloud organization. • Cloud coordination agent: in charge of the coordination of tasks, events, and information flows triggered by the user (events coming from the view served by the cloud visualization agent) and the system (events created by this agent in a cyclic way to perform system tasks), within the cloud virtual organization. • Cloud visualization agent: provides the user with a graphical interface to interact with the system. 3.3

Analysis Architecture

The proposed agent architecture is presented as a service architecture. Given that the layers, and their respective types of nodes have been defined, in this section the architecture’s components are defined. For this purpose, several components have been defined, as can be described as follows: – IoT physical nodes: consists of sensor nodes (composed of data aggregation probes, and physical sensors that perform physical measurements), responsible for collecting data from the environment. Subsequently, the nodes send said data periodically through the IoT wireless transmission system to the event management system. – IoT wireless transmission system: a network system responsible for providing wide-area coverage to physical IoT nodes in order to communicate with the event management system. It should be noted that this system has several proposed technologies as the basis for its implementation, which is chosen on the basis of the types of sensors, gateways, and other available physical devices.

44

S. M´ arquez-S´ anchez et al.

– Event management system: this is a component that works in the event-driven paradigm, collecting data from the producing subsystems and sending them to the consumer systems subscribed to it. – Data analysis and federated learning subsystem: receives all the data from the event management system, applies pre-processing and analysis to it at the edge. This analysis is part of the federated learning paradigm, since it is the method that best adapts, in this case, to physical deployment. Subsequently, the subsystem sends the data and results to the cloud ecosystem to allow for their usage by the different cloud services. – Artificial intelligence coordination service: this service is responsible for carrying out the heaviest statistical and artificial intelligence multiple analysis, which cannot be carried out in the edge environment. It is also be responsible for coordinating the federated learning system i.e., knowledge exchange between the edge nodes. – Task management service: carries out task management in the background. – Ingest management service: performs data ingestion and stores data in the persistence system. – Notification alert service: sends notifications to users and edge devices. – Web application: provides the user with the graphical interface which enables them to work with IoT data, analysis, etc. It is responsible for serving the functionality of the system to the user and allowing them to interact.

4 4.1

Case of Use Implementation Architecture

Finally, in accordance with the multi-agent and component architectures described in the previous sections, system implementation in the scope of a multi-agent system is illustrated in Fig. 3. It corresponds to the previous architecture, defining a possible stack of technologies to be used for this purpose. The technologies used for the implementation of the system are defined as follows: – IoT wireless transmission system: implemented through a Wi-Fi network which provides a large area of connectivity at the cost of reduced bandwidth, making it ideal for the use case that had been considered in this research. However, it may be necessary to use other technologies, such as LoRa, depending on the constraints of the physical environment of deployment and those of implementation. – Event management system: a free software system that implements the MQTT protocol [two], designed and built specifically for multi-sensor data ingestion use cases in IoT environments. Specifically, the selected system is eclipse Mosquito, under the eclipse license. – Nvidia Jetson: this is the edge node implementation. It is proposed to use an Nvidia Jetson board, to be able to carry out artificial intelligence analyses in the federated learning paradigm, developed in Python. Among other advantages, this language allows for agile development, helping to adapt the data

Adaptive and Intelligent Edge Computing Based Building

Fig. 3. Final implementation architecture

45

46









S. M´ arquez-S´ anchez et al.

to the needs of the components that belong to the IoT and cloud layers, since this component is the interface between these two. Artificial intelligence coordination service: It consists of an API rest, which contains integrated heavy artificial intelligence models that cannot be carried out at the edge, due to the computing demand they offer. It also holds the responsibilities of coordinating the process of exchanging knowledge regarding federated learning between the edge nodes. Since these are developed in Python, this API is also to be developed in Python together with Flask, the most widely used framework for developing API rest in Python today. Task service, ingest service, and alert notification service: are implemented in Python alongside Flask, the most widely used framework for API rest development in Python today, due to the flexibility they offer. Web application: a web application developed in JavaScript in the context of the MEVN stack. This component has two main sections: The backend, that is to be developed on the Node.js framework, Express due to the use of the MEVN stack. In the case of the frontend, the Vue.js framework is to be developed due to the use of the MEVN stack. Databases: It has been decided to use PostgreSQL, because it is the most used SQL database today, due to its extensibility, replication capacity and agile development ability. The outputs generated from this case study can be summarized as follows:

– A complete smart platform based on virtual organizations, designed as a 3-tier architecture able to ingest data from multiple sources and provide tailored responses for an efficient energy consumption pattern. – An IoT network and edge gateways able to satisfy the data ingestion needs of the platform to be deployed in the pilot stage, and to encrypt data at hardware level. – A data security and privacy protocol integrating cutting-edge cryptography solutions and a DLT-based approach. – A social machine able to manage the information processed by the platform, classifying and monitoring the information, identifying different scenarios and providing tailored responses (DSS). – New XAI approach (Deep Symbolic Learning) using hybrid neuro-symbolic artificial intelligence algorithms for a better integration of machine reasoning and learning capacities. – New predictive and optimization models based on hybrid symbolic learning. – Datasets gathered from the buildings involved in the demonstration phase, which enable the retrieval of information about consumption in buildings (anonymized/ ggregated and fully compliant with all ethic and privacy recommendations/legal framework). – Models (e.g., Energy consumption, suggestions vs. pattern modification, energy demand) considering social and human behavioral aspects (data correlation). The local platform allows users to consult the most up-to-date information collected by the sensors and the historical information collected by the sensors in

Adaptive and Intelligent Edge Computing Based Building

47

their home . The standard home automation protocols that allow communication with the sensors have been implemented. The local or edge system is responsible for sending all the necessary data to the Cloud on a regular basis. The system allows the user to consult, update, remove and register all existing actuators in the local environment, which can be controlled by the edge system. Moreover, the system makes it possible to visualize the relevant data, allowing the user to have a global vision of the actions and recommendations of the home.

5

Conclusion and Future Work

This work described a smart platform able to efficiently manage energy consumption in buildings (and districts), reducing consumption by means of suggestions that are aimed at a behavioral change in end users and power grid load optimization. To this end, the paper has reviewed the state of the art of literature on building and energy management systems (BEMS), generating multiple outputs. Future research will integrate all the developments of AI-BEMS into a usable platform to be tested in pilot tests. Real demonstration pilots will be built to test and validate the proposal. The designed architecture acts as a smart tool that promotes energy efficiency by contributing to a behavioral change in the user’s consumption pattern. This will result in a long-lasting energy consumption reduction. This system works not with one but with multiple aggregated users which means that a smart building energy consumption profile can be extracted and even extended to a district level, where the power grid benefits from a potential load shift, allowing for a proper distribution of the resources, making balanced use of the infrastructure and reducing, for example, overloads and maintenance costs. Acknowledgements. This research has been supported by the project “Adaptive and Intelligent Edge Computing based Building Energy Management System” Reference: NPRP13S-0128-200187, financed by Qatar National Research Fund (QNRF).

References 1. Lobaccaro, G., Carlucci, S., L¨ ofstr¨ om, E.: A review of systems and technologies for smart homes and smart grids. Energies 9(5), 348 (2016) 2. Sovacool, B.K., Furszyfer Del Rio, D.D.: Smart home technologies in Europe: A critical review of concepts, benefits, risks and policies. Renew. Sustain. Energy Rev. 120 (2020) 3. Orejon-Sanchez, R.D., Crespo-Garcia, D., Andres-Diaz, J.R., Gago-Calderon, A.: Smart cities’ development in Spain: a comparison of technical and social indicators with reference to European cities. Sustain. Cities Soc. 103828 (2022) 4. Ghayvat, H., Mukhopadhyay, S., Gui, X., Suryadevara, N.: WSN-and IOT-based smart homes and their extension to smart buildings. Sensors 15(5), 10350–10379 (2015)

48

S. M´ arquez-S´ anchez et al.

5. Shokri Gazafroudi, A., Soares, J., Fotouhi Ghazvini, M., Pinto, T., Vale, Z., Corchado, J.: Stochastic interval-based optimal offering model for residential energy management systems by household owners. Int. J. Electr. Power Energy Syst. 105, 201–219 (2019) 6. Gazafroudi, A.S., Soares, J., Ghazvini, M.A.F., Pinto, T., Vale, Z., Corchado, J.M.: Stochastic interval-based optimal offering model for residential energy management systems by household owners. Int. J. Electr. Power Energy Syst. 105, 201–219 (2019) ´ 7. Moreno, M.V., Ubeda, B., Skarmeta, A.F., Zamora, M.A.: How can we tackle energy efficiency in IoT based smart buildings? Sensors 14(6), 9582–9614 (2014) ´ Alonso, R.S., Prieto, J., Corchado, J.M.: Energy efficiency in public 8. Garc´ıa, O., buildings through context-aware social computing. Sensors 17(4), 826 (2017) 9. Minoli, D., Sohraby, K., Occhiogrosso, B.: IoT considerations, requirements, and architectures for smart buildings-energy optimization and next-generation building management systems. IEEE Internet Things J. 4(1), 269–283 (2017). https://doi. org/10.1109/JIOT.2017.2647881 10. Ganchev, I., Ji, Z., O’Droma, M. (2014). A Generic IoT Architecture for Smart Cities 11. Casado-Vara, R., Martin-del Rey, A., Affes, S., Prieto, J., Corchado, J.M.: IoT network slicing on virtual layers of homogeneous data for improved algorithm operation in smart buildings. Futur. Gener. Comput. Syst. 102, 965–977 (2020) 12. Metallidou, C.K., Psannis, K.E., Egyptiadou, E.A.: Energy efficiency in smart buildings: IoT approaches. IEEE Access 8, 63679–63699 (2020). https://doi.org/ 10.1109/ACCESS.2020.2984461 13. Solis-Mora, V.S., Gruezo-Valencia, D.F.: La Inteligencia Artificial (IA) al servicio de la eficiencia energ´etica en el Ecuador. Domino de las Ciencias 8(2), 600–621 (2022) 14. Mart´ınez, M., Santana, E., Beliz, N.: An´ alisis de los paradigmas de inteligencia artificial, para un modelo inteligente de gesti´ on de la energ´ıa el´ectrica. Revista de Iniciaci´ on Cient´ıfica 3(1), 77–84 (2017) 15. S´ anchez, A.E.G.: Optimizaci´ on de la operaci´ on de un sistema HVAC para ahorro energ´etico, mediante estrategias de Inteligencia Artificial (2020)

A Novel Aging-Based Proof of Stake Consensus Mechanism Mahmoud Abbasi1(B) , Javier Prieto2 , Marta Plaza-Hern´ andez1 , 1,2,3 and Juan Manuel Corchado 1

BISITE Research Group, University of Salamanca, Edificio Multiusos I+D+I, Calle Espejo 2, 37007 Salamanca, Spain {mahmoudabbasi,martaplaza,corchado,javierp}@usal.es 2 AIR Institute, IoT Digital Innovation Hub, Salamanca, Spain 3 Department of Electronics, Information and Communication, Faculty of Engineering, Osaka Institute of Technology, Osaka 535-8585, Japan

Abstract. Proof-of-Work (PoW) is currently the most popular consensus mechanism used in blockchain platforms. However, the significant amount of energy required by PoW has seriously undermined its effectiveness. Proof of Stake (PoS) is an alternative consensus mechanism that addresses the energy consumption issue of PoW. Unfortunately, the PoS mechanism may also lead to several issues, including a low level of decentralization, unfairness, and security issues in the blockchain. To address these concerns, we propose an aging-based version of PoS that enhances the decentralization level of the network by providing validation opportunities to a broader group of nodes. In addition to the nodes’ stake, aging-based PoS takes into account the age of nodes as a deciding factor in achieving fairness. Furthermore, aging-based PoS incorporates a punishment mechanism to prevent nodes from engaging in malicious activities on the blockchain. Our results demonstrate that aging-based PoS offers high decentralization, fairness, and security for current blockchain systems. Keywords: Blockchain mechanism

1

· Proof of work · Proof of stake · Consensus

Introduction

Bitcoin is a decentralized Peer-to-Peer network where nodes can transact with each other using the Bitcoin blockchain protocol [1]. Bitcoin uses a distributed consensus mechanism to track and verify transactions in the network. The consensus mechanism is the most crucial aspect of the blockchain system, as its performance directly impacts the blockchain’s operations, such as the accuracy and validity of blocks. Generally speaking, the consensus mechanism is an agreement between all the peers about which blockchain transactions are valid and c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 49–61, 2023. https://doi.org/10.1007/978-3-031-36957-5_5

50

M. Abbasi et al.

which are not. Proof-of-Work (PoW) is designed as the consensus mechanism for Bitcoin [2]. The idea behind PoW is that miners compete to generate blocks through solving a cryptographic puzzle, which requires notable computing and energy resources. A major criticism of PoW is that it consumes a considerable amount of electricity and, consequently, has negative environmental impacts [3]. To address the high energy consumption problem posed by the PoW mechanism, Proof of Stake (PoS) is proposed as an energy-efficient alternative [4]. In PoS, a single node is chosen from a group of nodes based on its stake (balance) to validate the next block. Compared with the PoW, the PoS mechanism requires low computational power, and there are no block rewards for validators. Despite the decisive advantages of PoS, it was noticed that there are some challenges with PoS-based blockchains. The first disadvantage of the basic version of PoS is an unfair chance of block validation. In other words, the PoS is in favor of rich nodes. This phenomenon is known as the “rich-get-richer” or the Matthew effect [5]. Such an unfair mining probability can eventually damage the decentralization feature of the network [6]. An additional problem with the PoS mechanism is the “nothing-at-stake” or “rational forking” problem [7]. In the nothing-at-stake, validators could create blocks on different chains (forks) with nothing at stake to double-spending attacks or reduce the system’s efficiency. Last but not least, the PoS is also vulnerable to 51% attack [8]. A 51% attack on a PoS-based blockchain refers to a situation where a validator or a group of validators would have more than 50% of the staked cryptocurrency. Validator(s) with such validation power can block new transactions from being conducted or confirmed. This paper proposes an aging-based version of Proof-of-Stake (PoS) consensus mechanism to address the challenges of decentralization, fairness, and security in blockchain systems. The proposed mechanism takes into account the age of nodes in addition to their stake, thus enhancing the validation opportunities of a broader group of nodes. Furthermore, the mechanism incorporates a punishment system to prevent malicious activities on the blockchain. The contribution of this work lies in its ability to improve the decentralization level of the network while ensuring fairness and security. Our proposed mechanism is a promising alternative to the widely used Proof-of-Work (PoW) and existing PoS mechanisms.

2

Related Works

Peercoin was the first cryptocurrency that implemented PoS as its consensus mechanism. Shortly thereafter, more cryptocurrencies were introduced that employed PoS-based consensus mechanisms, such as Nxt, BlackCoin, Cardano and Algorand. To date, different classifications of the PoS consensus mechanisms have been proposed [9]. In [10], the authors classified the PoS algorithms into four categories: committee-based PoS, chain-based PoS, DPoS, and Byzantine Fault Tolerant (BFT)-based PoS. Furthermore, they provided examples of cryptocurrencies that use each PoS class. For example, chain-based PoS is used by Peercoin

A Novel Aging-Based Proof of Stake Consensus Mechanism

51

and Nxt, or DPoS is implemented by BitShares 2.0 and Lisk. Another classification of PoS-based consensus mechanisms drawn by Saad et al. [11] divides them into randomized block selection and coin age-based selection. A combination of hit value and the highest stake is used in the former to select the next block validator. In the coin age-based selection method, a candidate with the maximum value of coin age is selected to validate the next block. In addition, researchers have introduced different variants of PoS to overcome the shortcomings of the initial version of the PoS mechanism, such as Delegated Proof of Stake (DPoS) [12], improved DPoS [13], DPoS with downgrade [14], DPoS with Dynamic Trust (DT-DPoS) [15], e-PoS [11], probabilistic PoS (LaKSA) [16,17], node preference-aware DPoS [18], game theory-based PoS [19], weighted voting PoS [20], and DPoS based on Vague Sets [21]. DPoS is a modified version of the PoS algorithm in which the nodes vote and select delegates or witnesses to validate the next block of transactions [12]. In DPoS, a node can vote on delegates by pooling its coins into a centralized staking pool and linking them to a specific delegate. A restricted number of delegates (usually between 20-100) are selected for each new block of transactions. When delegates are selected, they are responsible for deciding which transactions should be rejected and which one should be approved. Some works such as [13,14] try to overcome the weaknesses of DPoS by means of new ideas. In [13], a reputation-based DPoS idea has been introduced to alleviate the problem of the low enthusiasm of voting nodes and complications in dealing with malicious nodes. The work in [14] integrated the computational competition of PoW into DPoS in order to provide a consensus mechanism that is more efficient, fair, and decentralized. More specifically, in this mechanism, the influence of both computing power and stakes on creating new blocks is decreased to gain better performance. Overall, this section highlights the diversity and evolution of PoS-based consensus mechanisms, with different classifications and variants being proposed to improve their efficiency, fairness, and decentralization.

3 3.1

Methodology Problem Statement

Rich-get-richer issue is in favor rich validators and finally can break the decentralization of the PoS-based network when a few wealthy validators control the majority stake. The concentration of stake could have increased the likelihood of rollback attacks and the risk of data tampering. Carry outing rollback attacks require control of a notable portion of the total network stake to generate a longlived fork [22]. These types of attacks on blockchain systems cause significant risks. For example, if an adversary node changes the order in which transactions are conducted, the node can damage the system’s confidentiality. Another source of concern is the nothing at stake phenomenon. This phenomenon is explained by the assumption that in an initial versions of PoS, every validator will validate on all forks simultaneously when a fork takes place. One of the consequences of staking on multiple forks is disrupting consensus by delaying

52

M. Abbasi et al.

the consensus time and reducing the system’s efficiency. Moreover, it puts the network at risk of double spending attacks and other attacks. In blockchain systems, the 51% attack is also problematic when the PoS consensus mechanism is adopted, but it is improbable. 51% attack may happen when a malicious node in the blockchain controls more than 50% of the stake (or mining power) and exploits that majority to change the blockchain. The malicious node can prevent the confirmation of new transactions and change their order. This study aims to understand better the problems mentioned earlier and propose an effective PoS-based consensus mechanism to deal with or/and alleviate them. The aging technique will be used to increase the validation chance of the poor stakers and consequently increase the decentralization degree and fairness of the blockchain system (rich-get-richer and 51% attack). Moreover, a punishment mechanism is developed to prevent malicious activities (nothing at stake problem). 3.2

Proposed Method

This section introduces the proposed consensus mechanism called aging-based PoS. The phases for the aging-based PoS consensus mechanism are as follows. Phase 1: Initialization We first initiate aging-based PoS within a certain number of nodes and some selection rounds (e.g., 12 nodes and 200 rounds). Each node is associated with a stake, an age, and an address field. The stake field represents the number of coins the given node is staking in the blockchain. The age field shows a node’s age, starting from the first validation round in the blockchain system, and it is used to gradually increase the chance of the poor stakers (nodes) being selected as validator nodes based on their waiting time in the system. And finally, the address field represents the address of the node on the blockchain. Phase 2: Selection of consensus nodes The blockchain network comprises many nodes, and all nodes in this network are divided into regular and consensus nodes. In the blockchain system, the group of all nodes/peers is represented as N ode, and nodei denotes ith node in the blockchain. Furthermore, the stake and age of each node are represented as nodei .stake and nodei .age, respectively. We use the W innerP ool array to denote the group of all consensus nodes. To be a potential validator and placed in the W innerP ool array, a node must stake coins greater than zero (nodei .stake > 0). Phase 3: Calculating the total score achieved within the network During Phase 2, for each selected node, we calculate its score by multiplying the node’s number of staked coins and the node’s current age. Then, add the calculated score to T otalScore. Repeat this process for all the nodes in W innerP ool to find the total score achieved within the network (T otalScore).

A Novel Aging-Based Proof of Stake Consensus Mechanism

N odei .score = nodei .stake ∗ nodei .age T otalScore+ = N odei .score

∀nodes W innerP ool

53

(1)

Phase 4: Determining the winner node as the validator In this step, we select a node to be the validator of the next block through a function that will generate a random number (W innerN umber) between 0 and T otalScore. To give each node a fair chance at winning the selection process that is proportional to its score in the blockchain network, we gradually add the score of the current node (i.e., nodei .stake ∗ nodei .age) to T emporary variable. If W innerN umber is less than T emporary, that node is picked as the winner at any point. For example, assume that WinnerNumber is 46, and we have four consensus nodes that stake 12, 30, 18, and 45 tokens, respectively. Moreover, for the sake of simplicity, assume that the age for all the nodes is the same. If we gradually add the score of the current node and compare it with the winning number, we will have distributed the chance of winning into proportional sections. More specifically, the probability that the WinnerNumber place in the first 10-token section is 12/(12 + 30 + 18 + 45) = 12/105 = 11.4%. In other words, when a node has a big score (larger stake or greater age), its section will be larger, and thus the higher chance of winning. Phase 5: After selecting the winner Two situations may occur when we range over the W innerP ool array. In the first situation, the given node satisfies the conditional statement, i.e., W innerN umber < T emporary, and is selected as the validator of the next block. In this situation, to increase the chance of the other nodes (loser nodes) being selected as validators in the following rounds, we increase their ages by one unit. Remember that the node score is the multiplication of its stake in age. At the same time, to strengthen the incentive of the winner node to participate in the following validation rounds and also to make the validation process fairer, we increase its stake by ten units and decrease its age proportional to the profit obtained compared to its stake. For example, consider that the initial stake and age of node1 are 120 and 80, respectively. After winning a validation round, its new stake and age will be 130 and 70.4, respectively. The percentage decrease in its age is calculated as follows: Increase = N ew stake − Initial stake P ercentage increase = (Increase ∗ Initial stake)/100 Amount decrece = (P ercentage increase ∗ node1 .age)/100

(2)

N ode1 .age = node1 .age − Amount decrece In the second situation, the given node does not satisfy the conditional statement. In this case, we only increase the age of the current evaluated node with this logic that the current node has a low score (poor node) and consequently a low chance of being selected as a validator. Phase 6: The punishment mechanism The punishment mechanism is considered to prevent nodes from doing malicious activities, such as changing the block header hash and time conflict between the timestamp of the block header

54

M. Abbasi et al.

and the new block. The punishment process takes place before the consensus mechanism tries to add a candidate block as the new block to the blockchain. More specifically, the candidate block will be processed for validation to check if there are any errors in the blockchain. If the validation process finds an error, the offender node will be punished for malicious activity by losing its stake by ten units. If everything goes right, the candidate block will be appended to the blockchain as the new block. Note should be taken that to understand further the impact of different conditions on the performance of PoS, we also provide a modified version of aging-based PoS, in which when a node is selected for validation, it can not participate in the following validation process. In other words, this node excludes from the next validation round. This strategy can improve the decentralization degree of the initial PoS consensus mechanism over time and alleviate the richget-richer issue. We name this variant “conditional aging-based PoS.”

4

Simulation and Result Analysis

This section carefully analyzes our algorithm from the decentralization degree, fairness, and security perspective. Towards this end, we implement it in the Go programming language and design several experiments to analyze the behavior of aging-based PoS in different situations, such as regular and attack conditions. In our simulation experiment, we consider 12 nodes in the blockchain; however, setting the parameters (e.g., initial stakes and ages) for various experiments will be different to simulate real-world situations. More specifically, we consider three different groups of nodes, including poor, middle, and rich class nodes. For the poor nodes, the stake values are randomly generated from the range of 0–60; for the median and rich class nodes, this value is 64–127 and 128–200, respectively. Furthermore, the initial age value was randomly assigned for all nodes from 30–90. 4.1

Evaluation Parameters

The main measurement parameters used for our evaluation are described below. Jain’s fairness index: This index is important for evaluating system resources’ fairness and inequalities, and we adopted it for our work. Using this index, one can decide whether users obtain a fair percentage of system resources or not [23]. The Jain index equation is expressed as follows: n ( i=1 xi )2 (3) F I(x1 , x2 , ..., xn ) = n n. i=1 x2i where n ∈ N is the number of users (nodes), xi is the percentage of reward distribution for the ith node in a given round. The Jain index varies between (0, 1], where 0 is the worst case (unfair system) and 1 is the best case (maximum fairness).

A Novel Aging-Based Proof of Stake Consensus Mechanism

55

Load balance (LB) ratio: We use this measure to calculate the workload balance (reward distribution) among different nodes in a specific validation round. The Load balance ratio is denoted as follows: LB ratio =

The number of the round it has won during the number of the completed round ∗ 100 The number of the completed round

(4) In addition, we use the percentage of validated blocks for security analysis purposes to conduct our analysis in different conditions.

Fig. 1. Jain’s fairness, when the number of validation rounds is 200.

4.2

Fairness Analysis of Aging-Based PoS

Our simulation shows that the initial PoS algorithm is at serious risk of unfairness. This is mainly because this mechanism is in favor of the rich nodes. To deal with this challenge and achieve a reasonable level of fairness, aging-based PoS involves another factor, i.e., the node’s age, for the next block validator selection. Over time, the aging feature of the nodes increases the chance of the nodes with the low stake (poor stakers) to be selected as validators, consequently increasing fairness and reducing the risk of centralization. We can achieve this by modifying the stake and the age of nodes in each validation round. A line graph of the changes of the Jain’s fairness index generated by the 12-node size PoS, aging-based PoS, and conditional aging-based PoS consensus algorithms over 200 validation rounds is shown in Fig. 1. It is worth mentioning that we repeat all experiments five times and compute the average to increase the validity and reliability of our results. It can be seen in Fig. 1 that Jain’s fairness index provided by the initial PoS algorithm is lower than its counterparts. This means a lower level of fairness and decentralization. Moreover, it can be seen that the aging-based PoS and conditional aging-based PoS provide a reasonable level of fairness. Jain’s fairness

56

M. Abbasi et al.

index increases with time for both of them and reaches figures of 0.957 for agingbased PoS and 0.96 for conditional aging-based PoS at round 200. This is because we involve the node’s age and apply the conditional statement in the validator selection process. Figure 2 shows the behavior of three PoS consensus mechanisms during the validation process. Figure 2 displays the average number of validation rounds forged by each node compared to their initial stake in the standard PoS algorithm. Here, one can see that the black curve, i.e., the average number of rounds each node has forged, fluctuates wildly according to the node’s stake. This means that some nodes validated more fractions of blocks than others because they stake more coins (rich-get-richer problem). The Fig. 2 show the results for the aging-based PoS and conditional aging-based PoS algorithms, respectively. It is clear from these figures that the proposed algorithm (s) form a fairly smooth curve. In other words, the proposed algorithm (s) can alleviate the rich-getricher problem by a fair distribution of validation chance among different types of nodes (i.e., poor, median, and rich). As a result of this improvement, there will be a positive impact on the decentralization degree of the blockchain system and prevent rolled-back transaction attacks and data tampering risk.

Fig. 2. The validation behavior of different PoS algorithms with the random distribution of stake/age after 200 validation rounds

4.3

Decentralization Analysis of Aging-Based PoS

Our previous experiment investigated the blockchain system from a fairness perspective. Nevertheless, it would be interesting to know how the validation load distribution or reward distribution across the nodes would impact the system’s decentralization behavior. Toward this end, we conducted another experiment. We consider 12 nodes from different classes in terms of stake and age. Then we calculate the average percentage of workload balance or reward distribution for each node after each run (i.e., 200 validation rounds). We repeat this experiment for five runs and calculate the average. The results of the experiments are presented in Tables 1, 2 and 3. Table 1 shows the results for standard PoS algorithm. It can be seen that the average load ratio or reward distribution among different nodes is not balanced.

A Novel Aging-Based Proof of Stake Consensus Mechanism

57

In other words, the standard PoS algorithm is highly affected by the rich nodes. Over time, this can damage some key features of blockchain, i.e., decentralization and security. In particular, the node with a high stake would collect transaction fees and become a dominating validator (s) in such a situation. This increases the degree of centralization of the blockchain system and could lead to successful double spending attacks, compromising the system’s security. In contrast to standard PoS, the aging-based PoS algorithm uses an improved strategy to alleviate these problems while providing an appropriate level of reward distribution. Tables 2 and 3 show the results for aging-based PoS and conditional aging-based PoS, respectively. It is evident from the figures that the aging-based PoS algorithms achieve a relatively fair distribution of rewards among different nodes. This means that when we use aging-based PoS as the consensus mechanism in the blockchain system, it can increase the degree of decentralization because more nodes (especially the poor nodes) have a chance to select as a validator and an opportunity to participate in the decision-making process. Table 1. Experiment results: load balance ratio for standard PoS. Round # 200 Stake Run #1 Run #2 Run #3 Run #4 Run #5 Average

4.4

Node 1

60

2.5

7.5

8.5

7

4.5

6

Node 2

79

5

12

4.5

11

3.5

7.2

Node 3

102

8.5

18.5

13

7.5

15.5

12.6

Node 4

64

9

6

6.5

2

7

6.1

Node 5

87

8.5

9

11.5

5

7.5

8.3

Node 6

130

17

16

14.5

8

10.5

13.2

Node 7

57

4

7.5

2

4

8

5.1

Node 8

97

4

4.5

8.5

18

6.5

8.3

Node 9

117

15

8

16

13

15

13.4

Node 10

35

4

0.5

3

9

2.5

3.8

Node 11

70

16

3.5

4.5

4

5.5

6.7

Node 12

120

6.5

7

7.5

11.5

14

9.3

Security Analysis of Aging-Based PoS

One of the goals of this study is to decrease exposure to 51% attacks by making them highly improbable. To this end, we consider the potential attack scenarios to evaluate the validation behavior of the standard and aging-based PoS algorithm (s). More specifically, we consider the scenario in which a single dominating node has a significant stake that can play the role of the potential attacker. Table 4 presents the results for the standard PoS algorithm in five runs. In particular, the table shows the average percentage of the blocks generated by a single node with different dominant stakes, from 2000 to 12000 coins, where all other nodes keep the same stakes of 1000 coins. As can be seen, when the potential attacker has a relatively small stake, for example, 4000 coins or 26.66%

58

M. Abbasi et al. Table 2. Experiment results: load balance ratio for aging-based PoS. Round # 200 Stake Age Run #1 Run #2 Run #3 Run #4 Run #5 Average Node 1

60

37

8

7.5

7.5

7.5

6.5

7.4

Node 2

79

53

5

9.5

8

6

10

7.7

Node 3

102

77

8

10.5

8

9.5

5.5

8.3

Node 4

64

68

7.5

9

8

7.5

7.5

7.9

Node 5

87

68

8.5

10

11.5

9.5

8

9.5

Node 6

130

82

10.5

8.5

9

10

10.5

9.7

Node 7

57

80

9

7.5

9

7

9.5

8.4

Node 8

97

79

9

7.5

9

9.5

8

8.6

Node 9

117

79

10

9

8

10

9

9.2

Node 10

35

71

6.5

2.5

6

9.5

10

6.9

Node 11

70

64

9.5

9

8.5

8.5

8

8.7

Node 12

120

45

8.5

9.5

7.5

5.5

7.5

7.7

Table 3. Experiment results: load balance ratio for conditional aging-based PoS. Round # 200 Stake Age Run #1 Run #2 Run #3 Run #4 Run #5 Average Node 1

60

37

8

8

6

7.5

8

Node 2

79

53

9.5

6.5

7.5

9.5

10

7.5 8.6

Node 3

102

77

7.5

6.5

8.5

8

9

7.9

Node 4

64

68

6

7

8

7.5

5.5

6.8

Node 5

87

68

9

7.5

8.5

10

9

8.8

Node 6

130

82

10

8.5

12.5

12

8.5

10.3

Node 7

57

80

7

8

8

7

7.5

7.5

Node 8

97

79

8

10

7.5

6

8.5

8

Node 9

117

79

9

13.5

10

10

11

10.7

Node 10

35

71

8

6.5

4.5

6.5

6

6.3

Node 11

70

64

9

8

7.5

8

8

8

Node 12

120

45

9

10

11.5

8

9

9.5

of the entire network stake, the percentage of blocks created by the potential attacker is 28.3% of all the blocks. In this condition, the attacker cannot launch a 51% attack. However, when we assign a significant fraction of the stake to the potential attacker, for example, 12000 coins which correspond to 52.17% of the entire system stake, the system is vulnerable to 51% attack. This is because, in this situation, the attacker can create around 52.3% of all the blocks. The aging-based PoS algorithm (s) results are shown in the Tables 5 and 6. In this case, we assume the same scenario for these algorithms. When we use the aging-based PoS algorithm, the percentage of blocks created by the potential attacker when it has 12000 coins (i.e., 52.17% of the entire system stake) is only 5.9%. Hence, we can argue that the design of aging-based PoS is not vulnerable to this attack scenario (see Table 5). The conditional aging-based PoS algorithm performs well but is not as good as the aging-based PoS in this case (see Table 6).

A Novel Aging-Based Proof of Stake Consensus Mechanism

59

Table 4. Simulation results for standard PoS in a potential 51% attack scenario. 2000 (15.38%)

4000 (26.66%)

6000 (35.29%)

8000 (42.1%)

10000 (47.6%)

Run #1

18.5

23

34.5

42

46

12000 (52.17%) 52.5

Run #2

15.5

31

33.5

37

49.5

56.5

Run #3

16

28.5

34.5

46.5

48

45

Run #4

15.5

28

37

42.5

45.5

57

Run #5

18.5

31

32.5

41.5

48

50.5

Average

16.8

28.3

34.4

41.9

47.4

52.3

Table 5. Simulation results for aging-based PoS in a potential 51% attack scenario. Scenario

2000 (15.38%)

4000 (26.66%)

6000 (35.29%)

8000 (42.1%)

10000 (47.6%)

Run #1

9

10

8

7.5

6

12000 (52.17%) 4

Run #2

7.5

10.5

9.5

7

6.5

5.5

Run #3

8

9.5

10.5

6.5

5.5

5.5

Run #4

10.5

8.5

6.5

8

6

8.5

Run #5

8.5

9.5

8.5

8

8

6

Average

8.7

9.6

8.6

7.4

6.4

5.9

Table 6. Simulation results for conditional aging-based PoS in a potential 51% attack scenario. Scenario

2000 (15.38%)

4000 (26.66%)

6000 (35.29%)

8000 (42.1%)

10000 (47.6%)

12000 (52.17%)

Run #1

7.5

9.5

7.5

8.5

6.5

15

Run #2

7

9

8.5

8

5.5

14.5

Run #3

8

9.5

9

7.5

6

17

Run #4

7.5

9

8.5

8

6

17.5

Run #5

8.5

8

9

6.5

7

15

Average

7.7

9

8.5

7.7

6.2

15.8

5

Conclusion

In this paper, we first briefly introduce the standard PoS algorithm and highlight its pros and cons. The PoS algorithm suffers from network unfairness, centralization risk, and security vulnerability. In this study, we have provided an improved version of PoS, termed aging-based PoS, which ensures fairness and a high level of decentralization in the blockchain system, as well as resistance to 51% attack. Besides the stake, we introduce the age factor to be used in the validation process. For better performance of the proposed algorithm, it modifies these factors in each validation round. Moreover, we provide a punishment mechanism to prevent malicious activities. Through simulations and design of experiments, we show that aging-based PoS meets the objectives we defined at the first of the paper. In the future, we plan to deal with the storage overhead problem of the blockchain, mainly when one uses blockchain technology in applications with resource-constrained devices, such as the Internet of Things (IoT).

60

M. Abbasi et al.

Acknowledgement. This work was supported by the IoTalentum Project within the Framework of Marie Sklodowska-Curie Actions Innovative Training Networks (ITN)European Training Networks (ETN), which is funded by the European Union Horizon 2020 Research and Innovation Program under Grant 953442.

References 1. Nakamoto, S.: Bitcoin: A peer-to-peer electronic cash system. Decentralized Bus. Rev. 21260 (2008) 2. Jurdak, R., Corchado, J.M., Park, J.H., Bhatt, C., Dev, K.: Blockchain Based Sustainable, Secure Healthcare Systems, p. 109175 (2022) 3. Mezquita, Y., Gonz´ alez-Briones, A., Casado-Vara, R., Wolf, P., Prieta, F.D.I., GilGonz´ alez, A.-B.: Review of privacy preservation with blockchain technology in the context of smart cities. In: Sustainable Smart Cities and Territories International Conference, pp. 68–77. Springer, Berlin (2021) 4. Mezquita, Y., Parra-Dom´ınguez, J., P´erez-Pons, M.E., Prieto, J., Manuel Corchado, J.: Blockchain-based land registry platforms: A survey on their implementation and potential challenges. Logic J. IGPL (2022) 5. Liu, X.F., Jiang, X.-J., Liu, S.-H., Tse, C.K.: Knowledge discovery in cryptocurrency transactions: A survey. IEEE Access 9, 37 229–37 254 (2021) 6. Zhou, Q., Huang, H., Zheng, Z., Bian, J.: Solutions to scalability of blockchain: A survey. IEEE Access 8, 16 440–16 455 (2020) 7. Deirmentzoglou, E., Papakyriakopoulos, G., Patsakis, C.: A survey on long-range attacks for proof of stake protocols. IEEE Access 7, 28 712–28 725 (2019) 8. Aggarwal, A., Mezquita, Y., Valdeolmillos, D., Gupta, A., Gonz´ alez-Briones, A., Prieto, J., Corchado, E.S.: Blockchain module for securing data traffic of industrial production machinery on industrial platforms 4.0. In: Sustainable Smart Cities and Territories International Conference, pp. 39–47. Springer (2021) 9. Khobragade P., Turuk, A.K.: Blockchain Consensus Algorithms: A Survey (2022) 10. Xiao, Y., Zhang, N., Lou, W., Hou, Y.T.: A survey of distributed consensus protocols for blockchain networks. IEEE Commun. Surv. Tutor. 22(2), 1432–1465 (2020) 11. Saad, M., Qin, Z., Ren, K., Nyang, D., Mohaisen, D.: e-pos: Making proof-of-stake decentralized and fair. IEEE Trans. Parallel Distrib. Syst. 32(8), 1961–1973 (2021) 12. The bitshares blockchain. www.bitshares.org/technology/delegated-proof-ofstake-consensus/. Accessed 08 Aug 2022 13. Hu, Q., Yan, B., Han, Y., Yu, J.: An improved delegated proof of stake consensus algorithm. Procedia Comput. Sci. 187, 341–346 (2021) 14. Yang, F., Zhou, W., Wu, Q., Long, R., Xiong, N.N., Zhou, M.: Delegated proof of stake with downgrade: A secure and efficient blockchain consensus algorithm with downgrade mechanism. IEEE Access 7, 118 541–118 555 (2019) 15. Sun, Y., Yan, B., Yao, Y., Yu, J.: Dt-dpos: a delegated proof of stake consensus algorithm with dynamic trust. Procedia Comput. Sci. 187, 371–376 (2021) 16. Reijsbergen, D., Szalachowski, P., Ke, J., Li, Z., Zhou, J.: Laksa: A probabilistic proof-of-stake protocol (2020). arXiv:2006.01427 17. Dou, H., Yin, L., Lu, Y., Xu, J.: A probabilistic proof-of-stake protocol with fast confirmation. J. Inf. Secur. Appl. 68, 103268 (2022)

A Novel Aging-Based Proof of Stake Consensus Mechanism

61

18. Sun, Y., Yan, B., Yu, J., Duan, X.: A node preference-aware delegated proof of stake consensus algorithm with reward and punishment mechanism. In: International Conference on Wireless Algorithms, Systems, and Applications, pp. 572–583. Springer (2021) 19. Tyagi, N.K., Goyal, M., Kumar, A.: Game theory-based proof of stake mining in blockchain for sustainable energy efficiency. In: International Conference on Artificial Intelligence and Sustainable Engineering, pp. 121–132. Springer (2022) 20. Leonardos, S., Reijsbergen, D., Piliouras, G.: Weighted voting on the blockchain: Improving consensus in proof of stake protocols. Int. J. Netw. Manag. 30(5), e2093 (2020) 21. Xu, G., Liu, Y., Khan, P.W.: Improvement of the dpos consensus mechanism in blockchain based on vague sets. IEEE Trans. Ind. Inform. 16(6), 4252–4259 (2019) 22. Wang, Y., Cai, S., Lin, C., Chen, Z., Wang, T., Gao, Z., Zhou, C.: Study of blockchains’s consensus mechanism based on credit. IEEE Access 7, 10 224–10 231 (2019) 23. Lee, Y.L., Loo, J., Chuah, T.C.: Modeling and performance evaluation of resource allocation for lte femtocell networks. In: Modeling and Simulation of Computer Networks and Systems, pp. 683–716. Elsevier (2015)

Manizales, Smart and Sustainable Tourist Destination Luis Carlos Correa-Ortiz1(B) , Catalina Guevara-Giraldo1 and Elizabeth Chaparro Cañola2

,

1 Universidad de Manizales, Manizales, Caldas, Colombia {lcco,catalina.guevara}@umanizales.edu.co 2 IU Digital de Antioquia, Medellín, Antioquia, Colombia [email protected]

Abstract. The purpose of this article is to propose a model for Manizales, a city in the Andean Region of Colombia, as a Smart and Sustainable Tourist Destination (SSTD), through the development of the following objectives: establish the concept of Manizales as a smart and sustainable tourist destination through a review of various national and international standards and the formation of strategic work teams; establish the categories, strategic lines, and indicators of Manizales as SSTD; raise the baseline for the indicators and prioritize indicators obtained in light of the plan of development and other policies; and finally establish the action plan for Manizales as SSTD. Keywords: Smart and Sustainable Cities · Smart Destinations · Manizales · Key Performance Indicators

1 Introduction The Smart and Sustainable City (SSC) concept has become trendy given its relationship with sustainable development [1]. It emerged from what, in the late 1990s, was referred to as smart growth [2]. It is important to note that there is no consistent operationalization of what a smart and sustainable city is, as both experts in the area and academia do not have a precise conceptualization of such a city, even though its use has been increasing since the beginning of the new millennium [3]. An SSC is a city that monitors and integrates key elements of all critical infrastructure. This type of city, can optimize and make appropriate use of its resources, plan its preventive maintenance activities, and monitor safety aspects while creating value for its citizens [4]. A smart city is a city where all infrastructure is connected: physical, technological, social, and business institutional. The arrangement of this infrastructure facilitates and enhances the collective intelligence of the city, which has to do with making decisions that have an impact on a better quality of life for its inhabitants [5]. It is important to note that a smart city is one that combines Information and Communication Technologies (ICT) and web technologies with planning, design, and organizational efforts. This has the purpose of implementing efficient processes and identifying new and innovative solutions to manage © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 62–74, 2023. https://doi.org/10.1007/978-3-031-36957-5_6

Manizales, Smart and Sustainable Tourist Destination

63

the complexity of the city and ensure the improvement of its sustainability and livability [6]. Similarly, this type of city is characterized by the application of smart computing technologies in the management of the infrastructure and service components of a city [7]. However, achieving these city performance outcomes must involve respect for the principles of sustainable development [8]. The components of a smart and sustainable city are described below. Diverse perspectives on the components of an SSC have been found in the literature. These components make a city smart, efficient, and sustainable if integrated through ICT [9]. The key components of this type of city are soft and hard technology (soft-hard), people and institutions (government, strategies, and policies). The economic and human development of a city are a product of the relationship between these components [10]. The components of a smart and sustainable city that emphasize their relationship with aspects of urban life [11, 12] are presented in Table 1. Table 1. Components of a smart city and aspects related to urban life [11, 12]. Component

Aspects related to urban life

smart economy

industry

smart people

education

smart governance

e-democracy

smart mobility

logistics and infrastructure

smart environment

efficiency and sustainability

However, ICT corresponds to the key component that transforms a traditional city into a smart and sustainable one [9, 10]. Therefore, the components listed in Table 1 should be related to the information technology component [12]. The relationship between the key components of a sustainable city through this type of technology guarantees the improvement of social and human capital, which is an essential condition for achieving better economic performance. Additionally, it guarantees a better quality of life for citizens [13, 10]. Key Performance Indicators (KPI) represent a set of measures focused on those aspects of organizational performance that are most critical to the current and future success of the organization [14]. Therefore, in the SSC framework, KPIs correspond to indicators that serve to assess key outcomes to be monitored and managed in reference to the key components of these cities: the environment, non-ICT infrastructure, ICT infrastructure, public services, and human infrastructure [9, 11–13]. The convenience of the design, implementation, and analysis of indicators to evaluate the performance of a smart and sustainable city essentially address the possibility of improving the quality of life of its inhabitants [14]. In addition, KPIs allow comparisons in terms of the smart performance of one city with respect to another. This means that these indicators operate as a benchmarking mechanisms for SSCs [15]. Since tourism is included as a priority economic activity in the competitiveness agenda of Manizales [16], it is necessary to add this component to the SSC analysis. The

64

L. C. Correa-Ortiz et al.

application of the principles of SSC in tourism requires the interconnection of all actors in the value chain of this sector, through technological platforms where all data and information related to tourism activities are shared and managed. To achieve this, tourism must combine four elements: (1) ICT adapted to the management and control of tourism activities; (2) design and definition of data and information management processes from the client/tourist level and the institutional-organizational level; (3) availability of technological devices at multiple points of contact with the end user; and, (4) commitment of the actors in the tourism value chain to use the platforms in a dynamic manner [17]. Thus, the integration of such elements constitutes what is called a smart tourism destination [18]. The objective of installing these four elements in tourism is essentially related to improving the tourist experience, improving the performance of the tourism destination in reference to aspects that have to do with generating a competitive advantage [17].

2 Methodology The general objective of the project whose results are presented in this article is to propose a model for Manizales as a Smart and Sustainable Tourism Destination (SSTD). For this purpose, a qualitative methodology was proposed, specifically a case study with the following stages: first, a review of the concept of smart and sustainable tourism destination, and second, a review of at least four standards (two national and two international) according to the characteristics of the city. Subsequently, we established the categories, strategic lines and indicators that make up the model of Manizales as an SSTD, establishing the baselines for the indicators. Next, we prioritize the indicators obtained by considering the land use plan, the Manizales development plan, the revised standards, and through a focus group with stakeholders in the city, and finally establish the action plan for Manizales as an SSTD.

3 Revised Standards for Measuring Smart and Sustainable Tourist Cities In 2016, an alliance between the municipality of Manizales and the International Telecommunication Union (ITU) allowed the inclusion of the city in the United for Smart Sustainable Cities (U4SSC) program, a global UN initiative coordinated by ITU, UNECE, and UN-Habitat, which includes the ITU-T Y.4903 recommendation that provides KPIs for SSCs and establishes criteria to evaluate ICT contributions in making cities smarter and more sustainable [18]. Table 2 summarizes the dimensions and subdimensions of the KPIs. Based on the interest of the municipal administration of Manizales in defining tourism as the articulating axis of the city by conceiving it as intelligent and sustainable, it is necessary to include the UNE 178501:2018 standard in the revision. This standard provides a reference framework for the management, planning, and evaluation of the transformation towards intelligent tourist destinations [19]. Table 3 summarizes the strategic axes and descriptions of the standard. The third project included in this study was the initiative of Sustainable and Competitive Cities by Findeter in which Manizales participated in 2012. The Sustainable and

Manizales, Smart and Sustainable Tourist Destination

65

Table 2. Dimensions and subdimensions of the KPIs [22]. Dimension

Subdimension

Economy

ICT Productivity Infrastructure

Environment

Environment Energy

Sociocultural

Education, health, and culture Safety, housing, and social inclusion

Table 3. Strategic axes of UNE 178501:2018 standard [19]. Strategic axes

Description

Governance

The process of managing tourism destinations through the synergic and coordinated efforts of governments at different levels and with different attributions, the civil society living in the host communities and the business network related to the operation of the tourism system

Innovation

Innovative internal and external management approaches that translate into significant improvements oriented to the activities involved before, during and after the stay in the Smart Tourism Destination (STD), through the implementation of innovation management tools, competitive intelligence being one of them

Technology

Through the incorporation of technologies (information, communication, energy improvement, etc.) and technological monitoring that allows the use and application of data and content on markets, customers and products, an increase in the effectiveness and efficiency of the processes and services of the STD is pursued

Universal accessibility Universal accessibility and universal design as a way of adding value to all initiatives developed by STD stakeholders, taking as a starting point the strategies of awareness, training, and participation, with cross-cutting criteria and based on human diversity and equal opportunities. It extends to the entire tourism value chain: buildings, services, staff training, transportation, environments, web access, among others Sustainability

Sustainability contemplates the rational and efficient management of resources (environmental vector), the quality of life of tourists and residents (sociocultural) and business competitiveness linked to the economic vector

Competitive Cities initiative seeks to improve the quality of life of the inhabitants of

66

L. C. Correa-Ortiz et al.

Colombian cities and promote their sustainable development and competitiveness [20]. Table 4 presents the dimensions included in the initiative. Table 4. Dimensions covered in Findeter project [20]. Dimension

Description

Environmental Sustainability and Climate Change

Environmental management and local pollution control Greenhouse gas mitigation Energy efficiency and renewable energy Vulnerability reduction and adaptation to natural disasters

Urban Sustainability

Growth control and urban habitat improvement Urban equity

Economic and Social Sustainability

Sustainable urban mobility Local competitiveness Security and citizen coexistence

Fiscal sustainability and governance

Revenue management Expenditure management Governance of policy decision, planning, budgeting, and implementation processes

Finally, the historic center of Manizales obtained certification in 2019 as a Sustainable Tourism Destination under the technical standard NTS TS 001-1 by Icontec (Colombian Institute of Technical Standards and Certification), which was renewed in 2021. The dimensions of the standard are listed in Table 5 [21].

4 Manizales, Context and Vision of the City The model that Manizales defines to achieve this vision must be in accordance with the conditions of its own context, since most of the existing standards reviewed correspond to European or North American countries with very different social and economic conditions. Therefore, to define an adequate model, it is important to conduct a contextual review of Manizales to adapt the references to its own territorial conditions. The following is a general overview of Manizales in terms of quality of life, in numbers before (2019) and after (2021) the pandemic. Like almost all Colombian cities, Manizales has made progress in reducing poverty. In 2019, 11.9% of citizens were below the poverty line, which is equivalent to approximately 54 thousand citizens in this condition. In the last ten years, close to 90 thousand people have left poverty. The city has a Gini coefficient of 0.43, which, although it has

Manizales, Smart and Sustainable Tourist Destination

67

Table 5. Criteria contemplated in Icontec NTS TS 001-1 [21]. Dimension

Description

Environmental

Biodiversity protection Support for biodiversity protection and sustainable use programs Water management Energy management Use of chemical products Waste management Management of atmospheric, noise, and visual pollution Management of greenhouse gas emissions Management of emissions of ozone-depleting substances Management of other environmental impacts

Sociocultural

Management and prevention of social risks Support for communities Support for ethnic groups Satisfaction of resident populations Conservation, protection, and sustainable use of cultural heritage

Economic

Development of business capacity and job creation Organization of informal vendors Commercialization of inputs, goods, and services of the resident population Satisfaction of visitors and tourists Tourism product and promotion Economic monitoring

ceded 11% in the last decade, is still considered a high inequality according to the UN. In education, there is still important work to be done to improve performance and close gaps in educational quality, while 90% of private schools are in the A and A+ levels (the best performance categories); in official schools this proportion drops to 23%. In the environmental dimension, water and energy consumption has been reduced, and 76% of the measurement points of the Chinchina River reported regular, bad, or very bad water quality. However, progress was made in the wastewater treatment plant project, which has already been put out to bid. Solid waste production increased by 52% between 2010 and 2019 and the recycling ratio remains at 2,5%. By 2021, the production of solid waste will increase by an additional 23%, whereas the recycling rate will barely increase to 3.3%. In air quality, PM10 emissions were reduced between 2011 and 2019, and the only monitoring station for PM2, 5 registered a slight increase, a trend that continued through 2021. In mobility, by 2019 the accelerated growth of the vehicle fleet continued, which in the last decade increased by 132%, while public transport passengers decreased

68

L. C. Correa-Ortiz et al.

by 2% compared to 2018 and 17% compared to 2007. From the perspective of urban sustainability, Manizales is not moving in the right direction [22, 23]. Regarding the vision of the city, the main medium-term planning instrument available to Colombian territories is the Territorial Ordinance Plans (TOP). In these documents a planning of where the city is going and what the bets will be to achieve the proposed vision is made. It is important to frame the sectoral programs in the TOP so that their relevance to achieving a medium-term vision is visible. Manizales had a TOP approved in 2017 and was included in the analysis [24].

5 Model for Manizales as a Smart Sustainable Tourist Destination Thus, considering the context indicators, the vision of the TOP and its land use model, and articulating the elements established in the different international standards, the following model of Manizales as a Smart and Sustainable Tourist Destination (SSTD) was defined and is presented in Fig. 1.

Fig. 1. Model for Manizales as a Smart Sustainable Tourist Destination

Once the conceptual framework, review of standards, and model for Manizales were established, the next phase of the project consisted of setting up an indicator board to serve as a tool for monitoring the progress of Manizales as an SSTD. According to the characteristics of the SSTD, four major categories were established that grouped together the requirements that a city must meet to acquire this designation: environmental, economic, governance, and sociocultural. For each category, strategic lines and result indicators were defined as a guide so that the actions revolve towards the same objective: the positioning of Manizales as a smart and sustainable tourist destination. The indicator board has four categories, 21 strategic lines and 127 indicators. A summary of the battery indicators is presented in Table 6. After defining the battery of indicators,

Manizales, Smart and Sustainable Tourist Destination

69

data collection from different national and local sources began with 2016 as the baseline for the project. It is necessary to clarify whether all indicators included in the battery, whether quantitative or qualitative, are important for monitoring the impact of the project in the city. However, for some of them, there was insufficient information available at the time of project development, so these indicators were marked as pending. This leads to one of the general recommendations of the project: the centralization of information and access to open data. Table 6. Summary of the battery of indicators for Manizales as a SSTD. Category

Strategic Lines

General

General

4

Environmental (SDGs 6, 7, 11, 12, 13 & 15)

Water and Sanitation

6

Air Quality

4

Economic (SDGs 8, 9 & 11)

Number of Indicators

Energy

3

Public Space and Nature

3

Risk Management

7

Mitigation of Climate Change

3

Solid Waste

5

Noise

2

Competitiveness

12

Urban Habitat

5

Mobility

19

Public Services

7

ICT

5

Tourism

14

Governance (SDGs 16 & 17)

Public Finances

1

Public Management and Transparency

4

Sociocultural (SDGs 1, 2, 3, 4, 5 & 10)

Culture, Recreation and Sports

4

Education

5

Poverty and Inequality

2

Total

Health

3

Security

9 127

70

L. C. Correa-Ortiz et al.

6 Action Plan After the conceptual review, the definition of the model and its indicators, and the establishment of the baselines for these, the following actions are proposed that will allow the city to consolidate itself as an SSTD, which are framed within the categories defined in the model, which in turn respond to the expectations of international standards, the city context, and the vision that Manizales has set for 2032. In addition, they responded to the challenges prioritized in the expert validation roundtables with their public policy, relevance, and importance filters. The action plan is listed in Table 7.

7 Conclusion and Future Work The revision of the standards for measuring smart and sustainable tourism cities and their application in Manizales is a multi-dimensional process that requires the consideration of various factors. The UNE 178501:2019 standard for intelligent tourist destinations and the ITU-T Y.4903 recommendation for smart sustainable cities provide KPIs and criteria to evaluate ICT contributions towards making cities smarter and more sustainable. Additionally, the Sustainable and Competitive Cities initiative by Findeter aims to improve the quality of life of citizens and promote their sustainable development and competitiveness through financing and technical assistance to local governments for the implementation of projects in various sectors. The certification of the historic center of Manizales as a Sustainable Tourism Destination under the technical standard NTS TS 001-1 is evidence of the commitment of the city to sustainability in tourism. By considering these standards, initiatives, and certifications, Manizales can achieve the goal of defining tourism as the articulating axis of the city by conceiving it as intelligent and sustainable. A proposed model for Manizales as a Smart and Sustainable Tourist Destination (SSTD) is presented. The model considers context indicators, the vision of the TOP, its land use model, and international standards. The model includes four major categories— environmental, economic, governance, and sociocultural—with 21 strategic lines and 127 indicators. The next phase of the project will involve setting up an indicator board to monitor the progress of Manizales as an SSTD. The battery of indicators includes both quantitative and qualitative measures and is based on data collected from different national and local sources. It recommends the centralization of information and access to open data to improve the monitoring of the impact of the project on the city. Overall, this model provides a useful framework for Manizales to position the city as a smart and sustainable tourist destination. Finally, the proposed action plan for Manizales as a smart and sustainable tourist destination aims to address the dimensions included in the model. It includes specific actions such as the design and implementation of sustainable lodging seals, public air quality policy, mobility master plan, and establishment of a management entity for the tourism sector. If implemented effectively, these actions can help the city achieve the goal of becoming a smart and sustainable tourist destination, while also improving the quality of life for its residents.

Manizales, Smart and Sustainable Tourist Destination

71

Table 7. Action Plan for Manizales as SSTD. Dimension

Strategic Line

Prioritized Indicators

Key Actions

Environment

Water and Sanitation

Water consumption in tourist accommodations (liters/day)

Design and implementation of a sustainable lodging seal

Solid Waste

Urban solid waste generation per capita Solid waste generated by tourism operators

Economy

Air Quality

Existence, monitoring, and compliance with air quality regulations

Public Spaces and Nature

Square meters of effective Design and public space per capita implementation of the master plan for public space

Noise

Exposure to noise: percentage of residents exposed to excessive noise levels

A study on noise exposure by commune, disaggregated by types of noise

Mobility

Shared bicycles

Design and implementation of the mobility master plan

Bicycle lanes Number of vehicles per 1000 inhabitants

Design and implementation of the public air quality policy

Electric vehicle charging points Bicycle usage points Travel time index Economy

Competitiveness

Employment in the ICT industry Employment in the tourism industry

High-impact lodging: business development program to boost productivity in tourism sector companies (continued)

Acknowledgements. This project was executed within the framework of the partnership agreement 1906260525 between the ICT and Competitiveness Secretariat of the Municipality of Manizales and University of Manizales and would not have been possible without the collaboration of the following entities: Municipality of Manizales—Secretariat of Administrative Services and Secretariat of Planning-, Manizales Como Vamos, Tecnoparque Sena Nodo Manizales and Universidad de Manizales.

72

L. C. Correa-Ortiz et al. Table 7. (continued)

Dimension

Strategic Line

Prioritized Indicators

Key Actions

Urban Habitat

Existence of urban development strategies or documents and spatial planning at the city level

Implementation, evaluation, and monitoring of the Territorial Ordinance Plan (TOP)

ICT

Number of public Wi-Fi access points in the city

Ensure continuity of service in Wi-Fi points Double the number of Wi-Fi points in the city

Governance

Tourism

Existence of a SSTD Consolidation of a Management Entity or a management entity or Tourism Intelligence Unit Tourism Intelligence Unit

Public Finances

Existence of information systems for monitoring municipal management

Open data information services for monitoring municipal management

Culture, Recreation and Sports

Percentage of population participating in any cultural activity

Generation of cultural supply indicators

Security

Citizens who feel safe

Integrated security intelligence center

Public Management and Transparency Sociocultural

Percentage of the population who were victims of a crime in the last 12 months Homicide rate per 100,000 inhabitants Mortality rate from traffic accidents Emergency service response time

References 1. Su, Y., Fan, D.: Smart cities and sustainable development. Reg. Stud. 57(4), 722–738 (2023). https://doi.org/10.1080/00343404.2022.2106360 2. Deakin, M., Al Waer, H.: From intelligent to smart cities. Intell. Build. Int. 3(3), 140–152 (2011). https://doi.org/10.1080/17508975.2011.586671 3. Chourabi, H., Nam, T, Walker, S., Gil-Garcia, J.R., Mellouli, S., Nahon, K., Pardo, T.A., Scholl, H. J.: Understanding smart cities: an integrative framework. In: 45th Hawaii International Conference on System Sciences, pp. 2289–2297. IEEE (2012). https://doi.org/10. 1109/hicss.2012.615

Manizales, Smart and Sustainable Tourist Destination

73

4. Jia, Q.S., Panetto, H., Macchi, M., Siri, S., Weichhart, G., Xu, Z.: Control for smart systems: challenges and trends in smart cities. Annu. Rev. Control 53, 358–369 (2022). https://doi.org/ 10.1016/j.arcontrol.2022.04.010 5. Harrison, C., et al.: Foundations for smarter cities. IBM J. Res. Dev. 54(4), 1–16 (2010). https://doi.org/10.1147/jrd.2010.2048257 6. El Hendy, M., Atalla, S., Miniaoui, S., Daradkeh, M., Mansoor, W., Bin Hashim, K.F.: Hybrid approach for developing strategic ICT framework for smart cities—a case study of Dubai’s toll gates (Salik). Smart Cities 5(4), 1554–1573 (2022). https://doi.org/10.3390/smartcities5 040079 7. Javed, A.R., Shahzad, F., ur Rehman, S., Zikria, Y.B., Razzak, I., Jalil, Z., Xu, G.: Future smart cities requirements, emerging technologies, applications, challenges, and future aspects. Cities 129, 103794 (2022). https://doi.org/10.1016/j.cities.2022.103794 8. ICTs for a Sustainable World. https://www.itu.int/es/sustainable-world/Pages/default.aspx. Accessed 11 Mar 2023 9. Mohanty, S.P., Choppali, U., Kougianos, E.: Everything you wanted to know about smart cities: the internet of things is the backbone. IEEE Consum. Electron. Mag. 5(3), 60–70 (2016). https://doi.org/10.1109/mce.2016.2556879 10. Nam, T., Pardo, T.A.: Conceptualizing smart city with dimensions of technology, people, and Institutions. In: Proceedings of the 12th Annual International Digital Government Research Conference: Digital Government Innovation in Challenging Times, pp. 282–291. ACM (2011). https://doi.org/10.1145/2037556.2037602 11. Albino, V., Berardi, U., Dangelico, R.M.: Smart cities: definitions, dimensions, performance, and initiatives. J. Urban Technol. 22(1), 3–21 (2015). https://doi.org/10.1080/10630732.2014. 942092 12. Lombardi, P., Giordano, S., Farouh, H, Yousef, W.: Modelling the smart city performance. Innov. Eur. J. Soc. Sci. Res. 25(2), 137–149 (2012). https://doi.org/10.1080/13511610.2012. 660325 13. Anthopoulos, L.G.: Understanding the smart city domain: a literature review. In: RodríguezBolívar, M.P. (ed.) Transforming City Governments for Successful Smart Cities. PAIT, vol. 8, pp. 9–21. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-03167-5_2 14. Hara, M., Nagao, T., Hannoe, S., Nakamura, J.: New key performance indicators for a smart sustainable city. Sustainability 8(3) 206 (2016). https://doi.org/10.3390/su8030206 15. Anthopoulos, L.G., Janssen, M., Weerakkody, V.: Comparing smart cities with different modeling approaches. In: Proceedings of the 24th International Conference on World Wide Web, pp. 525–528. ACM (2015). https://doi.org/10.1145/2740908.2743920 16. Agenda de Competitividad de Manizales. https://manizales.gov.co/RecursosAlcaldia/201507 021501372940.pdf. Accessed 12 Mar 2023 17. Buhalis, D., Amaranggana, A.: Smart tourism destinations. In: Xiang, Z., Tussyadiah, I. (eds.) Information and communication technologies in tourism 2014, pp. 553–564. Springer, Cham (2013). https://doi.org/10.1007/978-3-319-03973-2_40 18. ITU-T Y.4903. https://www.itu.int/ITU-T/recommendations/rec.aspx?rec=12884. Accessed 11 Mar 2023 19. UNE 178501:2018. https://www.une.org/encuentra-tu-norma/busca-tu-norma/norma?c=N00 60239. Accessed 15 Mar 2023 20. Manizales Sostenible y Competitiva: Prioridades y Bases para el Plan de Acción. https://man izales.gov.co/RecursosAlcaldia/201505291420477385.pdf. Accessed 14 Mar 2023 21. Norma Técnica Sectorial Colombiana: Destinos Turísticos de Colombia. Requisitos de Sostenibilidad. (NTS-TS-001-1). https://www.mincit.gov.co/CMSPages/GetFile.aspx?guid= 28e59887-a6ce-404a-9675-848d52c696e4. Accessed 14 Mar 2023 22. Informe de Calidad de Vida Manizales (2019). https://manizalescomovamos.org/informescalidad/informe-de-calidad-de-vida-de-manizales-2019/. Accessed 14 Mar 2023

74

L. C. Correa-Ortiz et al.

23. Informe de Calidad de Vida Manizales (2021). https://manizalescomovamos.org/informescalidad/informe-de-calidad-de-vida-de-manizales-2021/. Accessed 14 Mar 2023 24. Plan de Ordenamiento Territorial 2017–2031. https://manizales.gov.co/transparencia-yacceso-a-la-informacion-publica/planeacion/plan-de-ordenamiento-territorial-2017-2031/. Accessed 20 Feb 2023

Development of an IoT Network for Urban Orchards in High Vulnerability Areas in Colombia Juan M. Núñez V1,1(B)

, Juan Manuel Corchado2

, and Diana M. Giraldo3

1 BISITE Research Group, University of Salamanca, Calle Espejo, S/N. Edificio Multiusos

I+D+I, 37007 Salamanca, Spain [email protected] 2 Bisite Research Group, University of Salamanca, Calle Espejo, S/N. Edificio Multiusos I+D+I, 37007 Salamanca, Spain [email protected] 3 Universidad Autónoma de Occidente, Cali, Colombia [email protected]

Abstract. This article presents the development of an (Internet of Things) network for the scalability of urban gardens in Cali, Colombia, with the aim of initiating and promoting the transition from traditional agriculture to smart and sovereign agriculture. The context of this study are the most vulnerable areas of the city of Cali, Colombia, where traditional and low-productivity agriculture is practiced. In this article, a multipurpose IoT architecture for transitory crops is validated, along with a methodology for knowledge transfer and social innovation. The designed strategy contributes to food sovereignty in vulnerable communities. The method responds to field research and validation, structured in three phases: (1) analysis of the context and identification of capacities, (2) transfer of scientific-technical knowledge, and (3) application of agile design methodologies and technological co-creation. Finally, a highly impactful result was obtained: the transformation of contexts and the empowerment of communities towards sovereignty and selfmanagement of food, towards the minimization of socio-environmental factors that affect agricultural production and maximize the quality of life of the population. Keywords: Internet of things · AgTech · Urban Gardens · Genetic Algorithms

1 Introduction Historically, peoples, countries, and civilizations sprung up from agriculture and it is agriculture that drove their economies. The evolution of agriculture has been very progressive from agronomic point of view, however, in some parts of the planet, the technological progress has been very slow. This is the case of Latin America, where most producers and farmers face a very large technological gap, and where traditional and artisanal processes are still practiced, failing to cover and obtain quality harvests or effectively combat the effects of climate change. Added to the problem is the forced © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 75–86, 2023. https://doi.org/10.1007/978-3-031-36957-5_7

76

J. M. Núñez V et al.

displacement of peasants to urban areas, due to the decreasing amount of land available for cultivation. When, they migrate to the city, they face a reality they have never experienced before, which includes a reduced living space, unsuitable agroclimatic conditions, and limited-arable land. Therefore, technical use of urban gardens is required and their promotion as agents of change and transformation in certain social environments, often vulnerable due to production weaknesses and low technical-scientific expertise in creating urban gardens. Community urban gardens can reduce food insecurity and serve as green spaces that alleviate social, economic, and cultural problems [1]. That is why it is of vital importance to create social and technological innovation strategies that integrate different agricultural systems or forms of cultivation, adapting to the different spaces available in the city, to prepare for the high demand for food and water projected for 2050 [2]. According to the United Nations, the world’s population will increase by 2 billion, leading to an estimated total of 11 billion people by the end of the century, generating concern about a possible food and water crisis if action is not taken from technological, environmental, and humanistic perspectives. Normally, the areas with the greatest vulnerability are the most densely populated, where the continuous expansion of the territory creates immediate and inherent pressure on food security [4]. In those areas, most agricultural production comes from external farms and almost never from their own or internal productions. Each urban territory has its agro-climatic and soil conditions, the use of traditional cultivation systems, hydroponics, aquaponics, and aeroponics, is encouraged according to [5], as it has shown advantages in cost reduction and increased productivity. Likewise, the use of urban garden networks mediated by IoT (Internet of Things) for information exchange and traceability. Based on the above, it can be identified that there are many variables and technologies that can enhance community urban gardens since the most complex variables are limited space, water resources, agro-climatic conditions, and connectivity. However, from a technological point of view, throughout the 90s and towards the beginning of the 21st century, there has been a need to progressively incorporate technologies in traditional and urban agriculture, due to frequent signs of global warming and different indicators pointing to the pollution of the Earth. As this concern grew, the amount of data from the primary sector, such as agriculture, rapidly increased [6]. The internet was the perfect pretext for large agricultural industries to incorporate LAN and WAN networks through enabling technologies such as wireless sensor networks (WSN) and various wireless protocols [7]. Over time, a significant number of wireless nodes and devices connected to the internet have been developed for multiple applications, health, industry, logistics, transportation, smart cities, etc. The main objective of Internet of Things (IoT) is to analyze and process data that comes from field hardware in the cloud [8, 9]. The IoT has made it possible to gather high volumes of data, also in green agriculture. The growth of IoT is very rapid. At the end of 2018, there were 22 billion devices connected to the internet worldwide, and it is estimated that by 2030 there will be approximately 50 billion devices [10]. This increase has a clear explanation: the world’s markets are constantly growing and they require connectivity of devices or objects to the cloud. It is estimated that the Internet of Things market grew from $7.2 billion in 2017 to $27.8 billion in 2022 [11]. Competitiveness and the need to analyze data intelligently are the

Development of an IoT Network for Urban Orchards

77

fundamental pillars of Industry 4.0, which seeks flexible and sustainable production over time [12]. Finally, this article is structured as follows: Sect. 2 presents the case study context, Sect. 3 describes the implementation of the IoT network, and Sect. 4 presents the results obtained from the deployment of the IoT network for urban gardens.

2 Context The most important actor when working with knowledge transfer methodologies is the community. This project worked with two localities called La Dolores (Palmira district) and El Poblado II (neighborhood in Santiago de Cali), located in the department of Valle del Cauca. La Dolores has a population of 1200 inhabitants, while El Poblado II, belonging to the 13th commune in the district of Agua Blanca in the southwest of the city of Santiago de Cali, Colombia, has 16,533 inhabitants. This locality is one of the most violent, historically inhabited by families displaced by the armed conflict. In the process, 14 young people between the ages of 8 and 19, most of them of African descent with primary education, participated. Almost 70% of them had not previously had contact with technology, while the remaining 30% had received subjects with content such as Scratch at school. They all claim to have internet at home. Most of the participants’ grandmothers or parents’ continue to cultivate or their grandparents were involved in family farming. Therefore, 7 community gardens have been identified in this context. Social innovation and social design play an important role in this process and invite the transformation of the structures that deprive people of basic necessities. When social innovation and social design arise from and for the community, it enables the people within that community to appropriate them knowledge they require and to create the dynamics for the development of new realities in their territories.

3 Implementation of IoT Network Transitory crops are products with a great supply in our country, occupying the second place in importance after coffee. Recently, large companies have been acquiring these technologies to improve production, but in Colombia, there are many community gardens that have problems with their harvests, therefore they have not broken the paradigm of moving from artisanal to automated harvesting; it should be clarified that in many cases this is due to the costs and lack of information on the subject. Climatic factors such as temperature, solar radiation, relative humidity, soil moisture, wind, among others, influence the performance of transitory crops, as they affect plant growth and physiological processes related to fruit formation. These factors also indirectly affect performance by increasing damage caused by pests and diseases. The recommended level of water or soil moisture is essential to maintain adequate nutrient levels, weed control, and pest and disease control. Transitory crops grow in warm and humid environments in which the insects and pests that damage the crop also proliferate. The following methodology has been considered to design the solution of the IoT educational platform in the garden. It can be seen in Fig. 1.

78

J. M. Núñez V et al.

Mapping and context requirements

Transfer of technological knowledge with the community

Development of the educational IoT platform

Tests and validation

Fig. 1. Project methodology.

It begins with the collection of on-site needs and mapping of the work context. Participants, members of the community, analyze the available inputs and materials to enhance and create the concept of the orchard. Design criteria such as access to the Internet or local network, electrical and water connection points, ease of use, and real-time data logging (remote monitoring of data and water control) are considered. Through participatory workshop designs, hardware programming courses based on Arduino, Raspberry Pi, were carried out to promote the use of code in open spaces for the optimization and better production of orchards. The hardware design consists of a Raspberry Pi 4, which acts as a service concentrator and for WIFI and Bluetooth communication. Since the Raspberry Pi 4 does not have analog-to-digital converters, a Raspberry Pi Pico was used in series to perform the function of the data acquisition module. For the node’s realization, five criteria were considered: low cost, portability, connectivity, low consumption, and precision of the analog-to-digital converter for data acquisition. Prior to selecting the microcontroller to implement in the node, the families of the most representative manufacturers in the development of technology for sensor networks, such as Atmel, Texas Instruments, Microchip, Telos, Arduino, Raspberry, STM, and Freescale were analyzed. This can be seen in Fig. 2.

Fig. 2. Hardware architecture (sensor node).

Low power consumption is an important part of the design, therefore, it is another selection criterion, since no information is stored in the nodes; having high memory capacity is not relevant. There is an SD backup memory that acts as a data logger for when there are difficulties in the IoT connection. The ZigBee protocol was selected for this project mainly for its low power consumption, which is very important in the applications considered in this project, as locating the nodes requires long intervals of time without being intervened. The 802.15.4 standard falls within the scope of low-speed personal area wireless networks. This type of network is characterized by its simplicity

Development of an IoT Network for Urban Orchards

79

and low cost, providing wireless connectivity in applications that require very low power consumption. Two criteria were considered for the selection of the network topology: (1) number of nodes distributed in the orchard, and (2) distance between the end nodes and the coordinator. For small crops, a star topology was selected, which only allows for communication between the end devices (routers) and the network coordinator. Connectivity is vital for this project because there are places in Latin America that still do not have access to the Internet. For this, two network architectures were designed: a LAN network and an IoT network.

Fig. 3. LAN network.

A local monitoring architecture was designed to avoid disconnection moments with the cloud and to allow all those responsible for the orchard to view the data and make decisions about the crops. It is also important to track the data for further productivity analysis. This can be seen in Fig. 3.

Fig. 4. Stack IoT.

In addition, an IoT protocol stack was designed to explain the weekly evolution of the crop at the agricultural education level. The my Devices Cayenne cloud and network protocols such as MQTT and WIFI are used to send data to the cloud. The IoT protocol stack figure can be seen in Fig. 4.

80

J. M. Núñez V et al.

4 Results and Contribution 4.1 Multiple Linear Regression Models Data collection is carried out between corresponding variables, obtaining 5760 data points for four months at a sampling frequency of every 30 min, i.e., 48 data points per day. Using the data, the correlation of the variables with respect to productivity is calculated. Multiple linear regression is used as an analytical technique, and the KNIME program is used to make the respective prediction and generate value for urban gardens.

ET ST HR SM

y = β2 x2 + β3 x3 + βn xn + ε

(1)

p = 400 + 3.2ET + 1.5HR − 2ST −−10.2SM

(2)

Environment Temperature. Soil Temperature. Relative Humidity. Soil Moisture.

Where y (productivity) is called the response or dependent variable, x is called the predictor or independent variable, and β is the slope of the regression line and ε is a random error, which is assumed to have a mean of 0 and a constant variance of σ2. One of the most important results is the main effects of each factor or variable on the average productivity. Next, the data collected from the first study location, La Dolores Cali, Valle (Table 1). Table 1. Average of collected data vs productivity (La Dolores) Lettuce Cultivation Experiments

ET (°C)

HR (%)

ST (°C)

SM (%)

Yield (pound)

1

28.15

59.10

27.23

73.12

25

2

27.00

58.55

24.88

87.62

18

3

28.37

60.71

25.32

80.99

27

4

26.37

55.30

26.30

79.30

17

5

29.00

61.00

27.88

83.50

20

6

27.35

62.12

27.30

78.52

22

According to the different repetitions and experiments, it was possible to obtain optimal ranges of values for the different environmental and soil variables to contribute to good productivity (Table 2). The same procedure was carried out in another georeferenced context such as El Poblado II, where the following average data were obtained. The same amount of data was collected for 4 months (Table 3 and Fig. 5).

Development of an IoT Network for Urban Orchards

81

Table 2. Range of variables for good productivity (La Dolores) Lettuce optimal ranges Variables

Orchard

Literature

Environment Temperature (°c)

28.50

26.00

Relative Humidity (%)

59.90

53.00

Soil Humidity (%)

85.30

77.00

Soil Temperature (|c)

28.38

27.00

Luminosity (% lx)

77

60

Table 3. Average of collected data vs productivity (El Poblado II) Lettuce Cultivation Experiments

ET (°C)

HR (%)

ST (°C)

SM (%)

Yield (pound)

1

27,47

73,93

28,73

79,33

23

2

28,39

78,73

26,5

70,17

13

3

28,83

83,81

25,01

78,19

16

4

29,29

82,25

26,1

78,72

27

5

28,25

65,61

26,7

84,25

24

6

27,92

69,9

27,77

78,59

24

Fig. 5. Data collection (La Dolores) and data collection (El Poblado II).

According to different repetitions and experiments, it was possible to obtain optimal ranges of values for different environmental and soil variables to contribute to good productivity. The results of the experiment indicate that the lettuce yield is influenced

82

J. M. Núñez V et al.

by the environmental variables of ET, HR, ST, and SM. In particular, experiments 1, 4, and 5 had the highest yields, suggesting that the environmental conditions in those experiments were optimal for lettuce cultivation. Lettuce is a crop that adapts well to a wide range of environmental conditions and is one of the most popular vegetables grown in urban gardens. In this experiment, environmental conditions were measured in urban gardens and lettuce yield was evaluated (Table 4). Table 4. Range of variables for good productivity (El Poblado II) Lettuce optimal ranges Variables

Orchard

Literature

Environment Temperature (°c)

30.15

26.00

Relative Humidity (%)

62.03

53.00

Soil Humidity (%)

83.14

77.00

Soil Temperature (|c)

29.00

27.00

Luminosity ( % lx)

85

60

4.2 Genetic Algorithms Genetic algorithms are optimization and search techniques based on principles of biological evolution. They can be used to find optimal solutions to optimization and search problems in different areas, including the correlation of environmental variables in urban gardens. For the genetic algorithm, a population of 100 individuals with 60 iterations was considered. This algorithm was implemented in Python, where libraries such as Pandas were used for data manipulation. Below is an example of Python code that implements a genetic algorithm to find the optimal ranges of input variables (ET, HR, ST, and SM) that maximize lettuce crop yield (Table 5). The table shows the optimal ranges found by the genetic algorithm for each variable in two different orchards, "La Dolores Orchard" and "El Poblado II". The literature for each variable is also provided as a benchmark for comparing the optimal ranges found. It can be observed that there are significant differences in the values obtained for each variable in both locations. In particular, the ambient temperature is slightly higher in La Dolores Orchard (29.89 °C) than in El Poblado II (29.98 °C), and in both cases it is higher than the literature temperature (26.00 °C). On the other hand, the relative air humidity is slightly lower in El Poblado II (58%) than in La Dolores Orchard (62%), but in both cases it is higher than the humidity in the literature (53%). Soil moisture is higher in both locations (81.69% in La Dolores Orchard and 83.81% in El Poblado II) than the moisture in the literature (77%). Finally, soil temperature is slightly higher in El Poblado II (29.00 °C) than in La Dolores Orchard (27.39 °C), and in both cases it is higher than the soil temperature in the literature (27.00 °C). Regarding the percentage of error, it can be observed that the variables with the highest error are ambient temperature and luminosity, with errors exceeding 15% in both locations. Soil

Development of an IoT Network for Urban Orchards

83

Table 5. Table of optimal ranges using genetic algorithms (La Dolores—El Poblado) Lettuce optimal ranges (genetic algorithms) Variables

La Dolores Orchard

ElPoblado II Orchard

Literature

% Error (La dolores Orchard)

% Error (El Poblado II Orchard)

Environment Temperature (°c)

29.89

29.98

26.00

15.35

15.31

Relative 62.00 Humidity (%)

58.00

53.00

16.98

9.43

Soil Humidity 81.69 (%)

83.81

77.00

6.41

8.99

Soil Temperature (|c)

27.39

29.00

27.00

1.44

7.41

Luminosity ( % lx)

81

76

60

35.00

26.67

moisture presents a lower error in both locations, with 6.41% in La Dolores Orchard and 8.99% in El Poblado II. These results indicate that, although different values have been obtained from the literature, there have not been significant deviations in the environmental conditions of the studied locations; Nevertheless, these are georeferenced optimal ranges for each population to minimize production losses. The behavior can be observed in the graphs in Fig. 6.

Fig. 6. Comparison of variables between orchards and literature.

84

J. M. Núñez V et al.

From the comparison of the graphs, it can be observed that both La Dolores Orchard and El Poblado II had higher values than the literature for all variables except soil temperature. The greatest differences between the orchards and the literature were observed for luminosity and relative humidity, with the orchards having those variables around 60% and 16–18% higher, respectively.

5 Conclusions It is important to transform the pre-established paradigm in which young people from these backgrounds perceive technical and mathematical learning as a component of little application in their daily lives. This is achieved through exercises that allow them to identify the usefulness of the knowledge acquired in the short term, supported by constant field accompaniment that is not only permeated by school activities but also by empathy dynamics and mutual knowledge generation. The horticultural communities need a high monitoring capacity to obtain information about their crops, analyze soil variability, and analyze soil and water quality. They also need to analyze their crop yields and productivity in relation to agroclimatic variables and soil variability. Learning in a real exercise allows the communities not only to appropriate knowledge and skills foreign to their contexts, but also to dignify the spaces and the image of these communities that are constantly stigmatized. In Cali, this area of the city is a consequence of the migrations of the armed conflict, many of these families were peasants from the Colombian Pacific and in their territories of origin they cultivated and lived on selfconsumption, therefore, with this type of project a connection with the origin and their ancestors is sought, which can be a great opportunity for the orchard technification exercise not only to be a technical process but also to reconnect and strengthen the knowledge of the families. In Colombia, areas like these exist in all cities, as it is an undeniable result of the great internal displacement that we have lived through for decades of war between different actors, which makes this type of practices known and easy to appropriate for many of the families living in the periphery and victims of the conflict. For some of the young people this may have been the first exercise where they produced food and for the first time took some kind of food home, even if it was just a few leaves of lettuce, rugula or chard, which gives a message of empowerment and hope to young people, which as teacher-researchers was satisfactory. The exercise of the garden mediated by technologies allows us to open to the community the possibility of thinking in new ways to mitigate hunger, stimulating the communities in the planting and harvesting of some of their food, but at the same time, it also allows the families to make visible and generate awareness regarding the use of water resources and the relationship between us and our environment and nature. In this type of process it is key to take into account the human capital and the real capacities of the communities, not only in terms of infrastructure but also human, such as job stability and income, level of education, free time available and support network, because without these aspects the project may not have the desired impact, since the technology could be implemented but the community would not end up making the appropriation that is desired.

Development of an IoT Network for Urban Orchards

85

Acknowledgments. This research has been supported by the project “Plataforma edge-iot basada en tecnologías dlt de alta eficiencia energética para el intercambio de tokens digitales verdes mediante la ejecución de contratos inteligentes, Reference: PDC2022-133161-C31, financed by MCIN/AEI/https://doi.org/10.13039/501100011033 and NextGeneration EU/PRTR, UE.”

References 1. Zhang, Y., Smith, J.P., Tong, D., Turner II, B.L.: Optimizing the co-benefits of food desert and urban heat mitigation through community garden planning. Landsc. Urban Plann. 226, 104488 (2022). ISSN 0169-2046 2. Nations, U., of Economic, D., Affairs, S., & Division, P. (n.d.). World Population Prospects 2019 Highlights 3. Cattivelli, V.: Pandemia e solidarieta’, l  analisi sulla citt‘ a di Cremona (2021). https://www. diocesidicremona.it/pandemia-e-solidarieta-lanal) 4. Pulighe, G., Lupia, F.: Food first: COVID-19 outbreak and cities lockdown a booster for a wider vision on urban agriculture. Sustain. 12, 10–13 (2020). https://doi.org/10.3390/su1212 5012 5. Gao, G., Xiao, K., Chen, M.: An intelligent IoT-based control and traceability system to forecast and maintain water quality in freshwater fish farms. Comput. Electron. Agric. 166, 105013 (2019). https://doi.org/10.1016/j.compag.2019.105013 6. Su, Y., Wang, X.: Innovation of agricultural economic management in the process of constructing smart agriculture by big data. Sustain. Comput. Informatics Syst. 31. https://doi.org/ 10.1016/j.suscom.2021.100579 7. J.M. Núñez, V., F. Fonthal R., Y.M. Quezada L.: Design and implementation of WSN for precision agriculture in white cabbage crops. In: 2017 IEEE XXIV International Conference on Electronics, Electrical Engineering and Computing (INTERCON), pp. 1–4 (2017). https:// doi.org/10.1109/INTERCON.2017.8079671 8. Laroui, M., Nour, B., Moungla, H., Cherif, M. A., Afifi, H., Guizani, M.: Edge and fog computing for IoT: a survey on current research activities & future directions. In: Computer Communications, Vol. 180, pp. 210–231 (2021). Elsevier B.V. https://doi.org/10.1016/j.com com.2021.09.003 9. Senyo, P.K., Addae, E., Boateng, R.: Cloud computing research: a review of research themes, frameworks, methods and future research directions. Int. J. Inf. Manage. 38(1), 128–139 (2018). https://doi.org/10.1016/j.ijinfomgt.2017.07.007 10. Number of internet of things (IoT) connected devices worldwide in 2018, 2025 and 2030, 2021. https://www.statista.com/statistics/802690/worldwide-connected-devicesby-access-technology/. Accesed 2 Jan 2022 11. IoT Analytics Market by Application (Energy Management, Predictive Maintenance & Asset Management, Inventory Management, Remote Monitoring), Component, Analytics Type, Deployment, Organization Size, & Vertical—Global Forecast to 2022, 2021. https:// www.marketsandmarkets.com/Market-Reports/iot-analytics-market-52329619.html#:~: text=The%20IoT%20Analytics%20market%20size,%2C%20industry%20vertical%2C% 20and%20region. Accesed 3 June 2022 12. Saravanan, G., Parkhe, S.S., Thakar, C. M., Kulkarni, V.V., Mishra, H.G., Gulothungan, G.: Implementation of IoT in production and manufacturing: an Industry 4.0 approach. Mater. Today Proc. (2021). https://doi.org/10.1016/j.matpr.2021.11.604 13. Sittón-Candanedo, I., Alonso, R.S., Corchado, J.M., Rodríguez-González, S., Casado-Vara, R.: A review of edge computing reference architectures and a new global edge proposal. Futur. Gener. Comput. Syst. 99, 278–294 (2019). https://doi.org/10.1016/j.future.2019.04.016

86

J. M. Núñez V et al.

14. Chen, C.H., Liu, C.T.: A 3.5-tier container-based edge computing architecture. Comput. Electr. Eng. 93 (2021). https://doi.org/10.1016/j.compeleceng.2021.107227 15. Maroli, A., Narwane, V.S., Gardas, B.B.: Applications of IoT for achieving sustainability in agricultural sector: a comprehensive review. J. Environ. Manage. 298. Academic Press (2021). https://doi.org/10.1016/j.jenvman.2021.113488 16. Lu, C., Grundy, S.: Urban agriculture and vertical farming. In: Encyclopedia of Sustainable Technologies, pp. 393–402. Elsevier (2017). https://doi.org/10.1016/B978-0-12-409548-9. 10184-8 17. AQUASTAT. http://www.fao.org/aquastat/en/overview/methodology/water-use. Accessed 6 Sep 2022 18. More people, more food, worse water. http://www.fao.org/3/ca0146en/CA0146EN.pdf. Accessed 6 Jun 2022 19. Liu, Y., Ma, X.Y., Shu, L., Hancke, G.P., Abu-Mahfouz, A.M.: From industry 4.0 to agriculture 4.0: Current status, enabling technologies, and research challenges. IEEE Trans. Ind. Inform. (2020). https://doi.org/10.1109/TII.2020.3003910 20. Nonaka, I.: A dynamic theory of organizational knowledge creation. Organ. Sci. 5(1), 14–37 (1994) 21. Spender, J.-C.: Making knowledge the basis of a dynamic theory of the firm. Strateg. Manage. J. 17(S2), 45–62 (1996) 22. Davenport, T.H., Prusak, L.: Working knowledge. Ubiquity 2000, 2 (2000) 23. Liyanage, C., Elhag, T., Ballal, T., Li, Q.: Knowledge communication and translation—a knowledge transfer model. J. Knowl. Manage. 13(3), 118–131 (2009) 24. Margolin, V.: Construir un mundo mejor—Diseño y responsabilidad social. Designio. México (2017) 25. Papanek, V.: Diseño para el mundo real. Ecología humana y cambio social. Academy Chicago Publishers, Chicago (1971) 26. Smith, A., Linder, B.: Manual Construcción de la Capacidad Creativa. Libro de diseño. MIT D-lab (2016) 27. J.M. Núñez, V., Vargas, V.L., and Y.M. Quezada, L.: Implementation of a participatory methodology based on STEAM for the transfer of ICT knowledge and creation of Agtech spaces for the co-design of solutions that contribute to the development of small and medium agricultural producers in Colombia, Panama and China. In: 2020 IEEE World Conference

DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture Gustavo Isaza1(B) , Fabian Ramirez1 , Néstor Duque2 Jeferson Arango Lopez1 , and José Montes2

,

1 University of Caldas, Manizales, Colombia

[email protected] 2 National University of Colombia, Manizales, Colombia

Abstract. Conventional techniques for the detection of distributed denial-of- service attacks have proven to be insufficient in the face of the variety and mutation of the typology of these anomalous behaviors, likewise, emerging detection and prevention technologies are presented as solutions in workstations and/or in processes not deployed as services. Therefore, this project implements an Intrusion Detection System (IDS) supported by deep learning techniques, with a serviceoriented architecture in the cloud for the detection of Distributed Denial of Service (DDoS) attacks, obtaining as a result the improvement of the classification metrics as well as the availability of these resources in an environment deployed as a service. Keywords: DDoS · Deep learning model using a cloud architecture · DDoS attacks detection · Deep neural network · Real-time detection DDoS and DoS

1 Introduction Over the years, there has been a significant increase in the number of cyber-attacks targeting home users, businesses, government organizations and even critical infrastructure. Where organizations that have a high percentage of Internet traffic are particularly vulnerable to DDoS attacks. This traffic can be generated from a variety of things, including online shopping, downloads and services. When a DDoS attack is perpetrated, normal traffic on servers is disrupted, causing a drop-in activity. DDoS attacks can also cause financial damage, as organizations are unable to provide certain services. In severe cases, DDoS attacks can even lead to the termination of business operations. For this reason, it is important to have automated systems capable of detecting intrusions reliably and responding quickly. In response to the above, a machine learning based DDoS (distributed denial of service) attack detection prototype with a cloud service oriented architecture using deep neural network is proposed by means of an implementation of parameter and hyper parameter defined networks that are tuned to each model based on the data provided in the dataset containing significant samples of DDoS attacks; The results of these intrusion detection techniques using deep neural networks materialise into promising advantages © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 87–96, 2023. https://doi.org/10.1007/978-3-031-36957-5_8

88

G. Isaza et al.

and actions that provide a real and feasible solution to the problem of intrusion detection and the wide variety of cyber-attacks, providing a better level of security through the proven identification of all malicious events observed in the network, by including all 79 characteristics of network traffic and analysing complex and difficult threats, as demonstrated by joint research experimenting with new and complex distributed denial-of-service attacks that are difficult for traditional approaches to clearly identify, outperforming signature and anomaly-based approaches that fall short in high-speed, high-data-flow networks.

2 Related Work Distributed denial of service (DDoS) is a modified denial of service (DoS) attack. According to [1] DoS attacks do not attempt to destroy or corrupt data, but to use a computer resource to the point where normal work cannot continue. Among the techniques for containing DoS, deep learning models have been proposed for detecting DoS attacks, [2] used a recurrent neural network (RNN) in the form of an automatic encoder to develop the deep learning model, evaluated their deep learning model using the CIC-DDoS2019 dataset and compared it with machine learning methods. On the other hand, [3] proposes a DDoS mitigation scheme for software-defined networks (SDN) that ensures accurate attack detection and efficient utilization of network resources. The scheme employs two stages: a bandwidth control mechanism and an extreme gradient boosting algorithm (XGBoost), evaluating the performance of the scheme using the CICDDoS2019, NSL-KDD, and CAIDA datasets. Among the methodologies [4] for detecting distributed reflection denial-of-service (DrDoS) attacks in IoT, two are presented: The first one uses a hybrid intrusion detection system (IDS) to detect IoT-DoS attacks. The second one uses deep learning models, based on short term memory (LSTM), expresses that the use of the proposed methodologies can detect misbehaviour making the IoT network safe from Dos and DDoS attacks. Pontes et al. [5] presented an energy-based flux classifier (EFC). This anomaly-based classifier uses inverse statistics to infer a statistical model based on labelled benign examples, given positive results obtained using different data sets. Akgun et al. [1] exposes DDoS attack detection using deep neural network (DNN) and short term memory (LSTM) algorithm, they propose a convolutional deep learning model based on inception type blocks, in the preprocessing section cleaning the redundant data as zero, nulls and duplicates are removed and extracted from the original CICDDoS2019 dataset, using different feature selection techniques on the obtained clean dataset, defining a subset of data with the information gain attribute evaluation technique. Zhou et al. [6] I design an optimal threshold model to filter suspicious flows, which reduces both computational and communication overheads, under a deep learning-based approach to design a new LSTM-based collaborative detection and prediction algorithm for collaborative DDoS attack prediction and attack flow detection in the environment, adapting a distributed publish/subscribe mechanism for efficient synchronisation of alert messages in large-scale distributed SDNs. Ganeshan et al. [7] Implements a Deep Neuro-Fuzzy Network based Fractional Anti Corona Virus Optimization (FACVO based DNFN) to detect DDoS in the cloud the

DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture

89

FACVO algorithm I use it to train DNFN in DDoS attack detection process, where the implemented FACVO algorithm incorporates both FC and ACVO using NSL-KDD dataset without attack and using BoT-IoT dataset. The proposed paradigm did not support real-time applications. The shortcomings of the above work are with respect to real-time and multi-class detection of DDoS and DoS attacks, as well as a complete take on all dataset features as previous research makes a significant reduction of these features.

3 Materials and Method For the design and execution of this research, the equipment described in Table 1, owned by Google, was available on a monthly subscription basis for use. A Google Colab Pro collaborative environment was also used, which allowed working with Jupyter Notebooks (see Table 2) and the Python programming language plus the database stored in the Drive. Implementation details are described in the github repository https://github.com/fab ian692/redes-neuronales-tesis. Where the high-level interfaces of Tensor Flow are under a standardisation of the Keras APIs for designing, training and testing neural networks, Keras enables the prototyping of neural networks in a more agile way; sklearn, pandas and numpy, support visualisation, data processing and application of classifiers pertaining to machine learning. Table 1. Hardware characteristics for design and implementation.

Processor

Product

Features

Intel Xeon

2.30 Ghz, 40 Cores, 40 Threads

Ram memory

38 Gb

Cloud storage

100 Gb

GPU

NVIDIA Tesla V100

16,16 Gb Vram, 5120 Cuda cores

TPU

V2-512

512 cores TPU, 4 TiB capacity

As for the process flow diagram, Fig. 1 shows the cicids 2017 dataset, data preprocessing, data normalization, set partitioning into 80%, 20%, the step through each deep layer and the outputs in predictions for each attack class. Additionally, metrics of classification results and ROC curves are calculated for each attack class. The neural network model has a 3 deep layer architecture, the first layer has 130 neurons with relu activation function, the second layer has 83 neurons with relu activation and the third layer has 78 neurons and Softmax activation function, which gives the predictions in the output as seen in Table 3 of the hyperparameters. According to the tensorgraph of the deep neural network model, the connection nodes are observed in addition to the relationships between the gradient, the optimizer, the size

90

G. Isaza et al. Table 2. Software features for design and implementation.

Python 3 Tensorflow 2.5.0 Scikit learn 0.22.2.post1 Pandas 1.1.5 Numpy 1.19.5 Keras 2.5.0 Google Colab PRO Ubuntu 18.04.5 LTS (Bionic Beaver)

Fig. 1. Process flow diagram.

of the neurons, the matching, the sequential model and the relationship of the predicted labels with the real labels; it should be noted that the highest number of tensors is in the path of the sequential Keras model, the gradient and the optimizer N adam. (see Fig. 2). 3.1 Data Set for Training The CICIDS2017 dataset proposed by [8] has updated attacks reflecting authentic realworld data, attacks in various categories and benign data in (PCAP) format. It also contains the results of network traffic analysis using CICFlowMeter, the attacks used are DoS and DDoS, with new instances in these categories with protocol filtering. This dataset meets 11 criteria that are characteristic in building a dataset that is a reliable reference: Complete network configuration, complete traffics, labelled dataset, complete LAN interaction and Internet communication, completeness of captured data,

DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture

91

Table 3. Hyperparameters 3 hidden layers. Hyperparameter

Adjustment

Hidden layers

3

Hidden layer activation function

Nadam

Output layer activation function

Softmax

Optimiser

Nadam

Learning rate

0.001

Bach size

200

Epochs

30

Loss function

Categorical sparse cross entropy

Fig. 2. Tensorgraf deep neural network.

current protocols, variety of attacks, heterogeneous data in conjunction with 78 traffic network characteristics, metadata. 3.2 Processing of the Dataset To process the dataset, the cicids 2017 daset is imported in 8 files in comma separated format, each file is filtered indicating the benign category, in file number 8 it is indicated to show all ddos categories except Heartbleed which does not belong to this category, they are concatenated and consequently a matrix with the grouped data is obtained with

92

G. Isaza et al.

a size of 2653785 columns and 79 rows, a repeated column with the name Fwd Header Length.1 is removed. After this step, new data for each class of DDoS attacks are added to improve detection and enforcement in real time because in practice with the standard cicids 2017 dataset when evaluated in real time the metrics decrease considerably, for this each attack category was carefully analysed with the tools Wireshark and etherape showing the traffic, After this analysis it was necessary to apply filters in Wireshark in each category of DDoS attacks to avoid the noise present in the network and only concentrate on the http protocol port 80 as this is the one affected in a denial of service, below is mentioned each filter applied by category. Re-index to avoid double-tagged rows, replace positive infinite and negative infinite numbers with the value zero (0), remove data with missing values in any row. Define the labelling of y as the data matrix of X, pre-process the data with the normalisation function, switch from labels to Label Encoder encoding to help normalise the labels so that they only contain values between 0 and n_classes-1 and group the data by attack and its quantity. The dataset partitioning of X_training, X_test, y_training, y_test is done with 80% partitioning on the data for training as 20% for testing. Preprocessing is necessary before training the proposed model, it removes outliers and scales the features to an equivalent range; normalisation was used in order to compare component sets and the mean by removing the effects of influences, the statistic is the scaling transformation of the distribution of a variable, which ensures a fast convergence of the learning process. 3.3 Validation with Real-Time Traffic As for the validation with real-time traffic, the data captured from the victim machine with the tcpdump tool is taken for each category of denial of service attack launched against the victim, capturing a file in.pcac format, which is uploaded to the IDS prototype deployed in the cloud, transforming the.pcac file to.csv with the cicflowmeter tool, which is sent to the deep neural network, responsible for making the prediction on the data sent. Where the predicted data is analysed with the malicious denial of service traffic that was launched and the number of files for each category is compared, where the two proposed algorithms give a promising response against the effective denial of service attack launched in the categories of denial of service and distributed denial of service in addition to the benign traffic generated in all the communication of the router with the rest of the computers performing common web browsing. 3.4 Cloud Architecture Registered users send data in .pcap format captured with tcpdump as shown in Fig. 3, the file is selected and sent, this file is sent to the prototype deployed on a server at the National University of Colombia, this makes the conversion from .pcap to csv with the cicflowmeter tool, this new file is sent to the deep neural network and returns the predictions for each category.

DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture

93

Fig. 3. Cloud architecture.

4 Results

Table 4. Metric Deep neural network classifier in training phase. Accuracy

Precision

F1 Score

Recall

0.9887

0.9893

0.99

0.9887

Table 5. Metric Deep neural network classifier in test phase. Accuracy

Precision

F1 Score

Recall

0.9886

0.9892

0.99

0.9886

As shown in Tables 4 and 5, the deep neural network in training phase and testing phase achieves good metrics of accuracy of 98.86%, precision 98.92%, f1 score of 0.99 and sensitivity of 98.86%, which indicates that the deep neural network model obtains good prediction response against the real label, which is very important in intrusion classification and detection of distributed denial of service attacks. As can be seen in Table 6 of the metrics the deep neural network obtained promising results, superior metrics in the training phase for all classes of denial of service attacks, which shows that the model is the most appropriate one to training. Furthermore, in the deep neural network metric, promising results are obtained in the highest accuracy score metric of accuracy in the test phase for all classes of denial-ofservice attacks, showing that the model is the most appropriate one to test (see Table 7).

94

G. Isaza et al. Table 6. Deep neural network metrics with training data filtering. Precision

Recall

F1-score

Support

BENIGN

0.9970

0.9949

0.9960

1830889

DDoS

0.9895

0.9923

0.9909

217046

DoS Golden Eye

0.9465

0.9703

0.9583

32319

DoS Hulk

0.9702

0.9833

0.9767

231026

DoS Slow httptest

0.7051

0.9095

0.7944

20589

DoS slowloris

0.9291

0.7741

0.8445

40396

0.9887

2372265

Macro avg

0.9229

0.9374

0.9268

2372265

Weighted avg

0.9893

0.9887

0.9888

2372265

Accuracy

Table 7. Deep neural network metrics with data filtering in test phase.

BENIGN

Precision

Recall

F1-score

Support

0.9969

0.9950

0.9960

456983

DDoS

0.9906

0.9925

0.9916

54821

DoS Golden Eye

0.9500

0.9710

0.9604

8058

DoS Hulk

0.9701

0.9834

0.9767

57933

DoS Slow httptest

0.7039

0.9039

0.7915

5192

DoS slowloris

0.9251

0.7695

0.8402

10080

0.9887

593067

Accuracy Macro avg

0.9228

0.9359

0.9260

593067

Weighted avg

0.9893

0.9887

0.9887

593067

Thus, the deep neural network model is very stable achieving accuracy at very close values in training and testing which is what is sought in an ideal model (see Fig. 4), that the lag of one with respect to the other is as small as possible and without ripples in the course of training in each epoch. Benign clase 0, Ddos clase 1, Dos Goldenye clase 2, Dos Hulk clase 3, Dos Slowhttptest clase 4, Dos Slowloris clase 5. The ROC curve area shows a deep model stability on the true positive rate axis, proving to be a good application of the model for the proposed solution for detection of denial of service attacks deployed in the cloud, roc curve of 1 indicating that it is an effective model in multi-class detection in DoS and DDoS detection (see Fig. 5).

DDoS Attacks Detection with Deep Learning Model Using a Cloud Architecture

95

Fig. 4. Accuracy model as a function of training and test phase times—loss model as a function of training and test phase times.

Fig. 5. Deep Neural Network Roc curve in test phase.

5 Conclusions According to the implementation of the prototype for DDoS (distributed denial of service) attack detection based on machine learning with service-oriented architecture in the cloud, 4 prototypes were made, where the 3 dense layer prototype reaches an ideal point in loss models and accuracy models in both training and testing phase, which is the symmetry and closeness of the two phases of training and testing and achieved the overall point of gradient descent that provides a deep neural network model Robust, efficient, agile and with a high level of intrusion detection metrics, far superior to the state of the art report, in addition, the correct detection was achieved for each of the 5 classes of distributed denial of service with its variants in the respective category plus the category of benign traffic.

96

G. Isaza et al.

The results are confirmed and consolidated in the joint validation between the National University of Colombia and the University of Caldas, where the model demonstrates that it is stable, fast, and highly accurate in detecting denial of service attacks in real-time, since real-time detection increases complexity, ratifying that the deep neural network model is agile giving predictions at high traffic volumes, stable and adjusts to unbalanced data and high dimensionality of the characteristics, which are those required for real-time detection where the effectiveness of the application of protocol filters for the improvement of the metrics is evidenced.

6 Limitations and Future Work When datasets are very large there could be data parallelization, which compromises response time and integrates other types of attacks. Acknowledgments. This work was carried out in the framework of the Project Prototype for DDoS attack detection based on machine learning in a cloud service-oriented architecture, with code 54308; approved in the framework of the Joint Call for Research, technological development and Innovation—2020, of the National University of Colombia and University of Caldas.

References 1. Akgun, D., Hizal, S., Cavusoglu, U.: A new DDoS attacks intrusion detection model based on deep learning for cybersecurity. Comput. Secur. 118, 102748 (2022) 2. Elsayed, M.S., Le-Khac, N.A., Dev, S., Jurcut, A.D.: DDoSNet: a deep-learning model for detecting network attacks. In: 2020 IEEE 21st International Symposium on “A World of Wireless, Mobile and Multimedia Networks” (WoWMoM), pp. 391–396 (2020) 3. Alamri, H.A., Thayananthan, V.: Bandwidth control mechanism and extreme gradient boosting algorithm for protecting software-defined networks against DDoS attacks. IEEE Access 8, 194269–194288 (2020) 4. Shurman, M.M., Khrais, R., Yateem, A.A.: DoS and DDoS attack detection using deep learning and IDS. Int. Arab J. Inf. Technol. 17, 655–661 (2020) 5. Pontes, C.F.T., de Souza, M.M.C., Gondim, J.J.C., Bishop, M., Marotta, M.A.: A new method for flow-based network intrusion detection using the inverse potts model. IEEE Trans. Netw. Serv. Manage. 18, 1125–1136 (2019) 6. Zhou, H., Zheng, Y., Jia, X., Shu, J.: Collaborative prediction and detection of DDoS attacks in edge computing: a deep learning-based approach with distributed SDN. Comput. Netw. 225, 109642 (2023) 7. Ganeshan, E.S.G.S.R, R., Jingle, I.D.J., Ananth, J.P.: FACVO-DNFN: deep learning-based feature fusion and distributed denial of service attack detection in cloud computing. Knowl.Based Syst. 261, 110132 (2023) 8. Sharafaldin, I., Lashkari, A.H., Ghorbani, A.A.: Intrusion Detection Evaluation Dataset (CICIDS2017) (2018). https://www.unb.ca/cic/datasets/ids-2017.html

Smart Cities Using Crowdsensing and Geoferenced Notifications Rui Miranda1 , Eduarda Ribeiro1 , Dalila Dur˜ aes1 , Hugo Peixoto1 , 2 1 onio Abelha , and Jos´e Machado1(B) Ricardo Machado , Ant´ 1

ALGORITMI Centre, University of Minho, Braga, Portugal [email protected],[email protected] {dad,hpeixoto,abelha,jmac}@di.uminho.pt,[email protected] 2 Cˆ amara Municipal de Guimar˜ aes, Guimar˜ aes, Portugal

Abstract. As the internet and the Internet of Things continue to expand, the idea of Smart Cities has begun to take hold. Smart Cities use connected devices and data to improve the environment and quality of life of their citizens. Technologies such as crowdsensing and geofencing allow citizens to contribute to initiatives and receive notifications when near areas of interest, respectively. This paper presents a systematic review of past works on the implementation of crowdsensing and geofencing technologies in Smart Cities, with the goal of identifying their purpose, strategies, and tools. The review examines seventeen relevant papers identified through the Scopus citation and abstract database. Keywords: Internet of Things · Smart Cities · Geofences · Crowdsensing · Crowdsourcing · Smart Notifications · Location-based Notifications

1

Introduction

As urban populations grow, the need to modernize cities and give essential services to communities has become increasingly important. This has led to the rise of Smart Cities, where technology and infrastructure are combined to enhance the quality of life in a sustainable and transparent way [1,11,23]. Smart Cities are increasingly dependent on the Internet of Things and its domains, as these technologies evolve and gain attention, from not only the academic community but also from industry and civil societies [14]. For many years, Smart Cities has been an area of interest and progress, but the initial phase of research was focused purely on its technical aspects. Emphasis has recently widened to include a more comprehensive approach to the information systems that support Smart Cities [13]. Nowadays, the focus is on enhancing citizen quality of life and assessing how smart technologies affect it, as well as on the social, economic, and environmental sustainability of cities [10]. One of the areas that has seen particular development is crowdsensing, as it enables to take advantage of the large amount of personal mobile devices and c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 97–110, 2023. https://doi.org/10.1007/978-3-031-36957-5_9

98

R. Miranda et al.

their sensors to collect and process information in a distributed and collaborative way, without more burden to the management or administrative entities [22]. Crowdsensing is a highly diverse field that has seen large growth and change. One aspect of its development involves integrating the sensory data collection through users’ mobile devices, allowing citizens to contribute to the collective via their devices. Another aspect is providing personalized, intelligent information to each citizen, enabling them to leverage the collective [16]. Regarding citizens taking advantage of the collective, some work and developments of interest have begun to appear most recently, with key technology for most these works being geofences, which consist of the notion of a virtual perimeter for a real geographical region. Devices and applications can use geofences to give useful information to citizens when they’re near an area of interest [8]. The implementation of geofences for the incorporation of citizens into intelligent environments through personalized information that’s sensitive to each citizen’s context, either through a mobile application or notifications, are goals for both the development and implementation of Smart Cities, as well as the expansion of crowdsensing users. The goal of this systematic review is to identify recent scientific publications that provide an overview of the current state of development in the areas of crowdsensing in smart cities. To this end, three research question were formulated: – RQ1: What is the role and purpose of crowdsensing and crowdsourcing, namely geofences and location-based notifications, in smart cities? – RQ2: What existing strategies and tools were utilized to incorporate crowdsensing and crowdsourcing into smart cities? – RQ3: How can crowdsensing implementations improve citizens’ lives in a smart city, and how make them active contributors? The following document is structured in five Sections: Introduction, Methodology, Results, Discussion, and Conclusion. This section’s goal is to present a contextualization and framing of the work, the main motivation, main objectives, and the document’s structure. Section two describes the research and review process, taking into account its main steps, i.e., selection of data sources, search strategy, selection criteria, and results. On Sects. 3 and 4, the obtained results are discussed, as well as relevant articles and documents are presented. The last section of this document aims to summarise and present the main conclusions and contributions obtained through the review, including recommendations for future work.

Smart Cities Using Crowdsensing and Geoferenced Notifications

2

99

Methodology

This research approach was guided by the PRISMA1 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement and checklist, widely accepted by the scientific community in the fields of computer science and engineering [15]. The steps taken into account include: 1. 2. 3. 4. 5.

Identification of the study’s research questions and relevant keywords; Creation of the research query; Definition of the eligibility criteria to filter and reduce the articles sample; Analysis of the resultant set of studies and papers; Presentation and discussion of the results.

On February 8, 2023, preliminary research was conducted using Scopus2 from Elsevier as the data source. The reasons for this choice were Scopus’ large abstract and citation database, quality assurance, and avanced search capabilites. To carry out the bibliographic research, keywords were identified and organized into two groups which were combined with a conjunction. To further refine the search, keywords in each group were combined with a disjunction. This ensured that documents containing at least one keyword from each group were selected, thus achieving the desired outcome. The research topic of ”Smart Cities”, ”Crowdsensing”, ”Smart Notifications” and ”Location-based Notifications” comprises the first group of areas and technical subjects, while the second group focuses on broader areas of technology such as ”Information Communication Technology”, ”Information Systems” and ”Mobile Computing” in order to better pinpoint results within the context of information systems and agents. As a result of applying the strategy described above, the following research query emerged:

Some eligibility criteria were defined in the form of exclusion criteria to screen the articles and studies collected, and all documents that met any of these criteria were excluded: 1. Publications that aren’t accessible in Open Access. 1 2

http://www.prisma-statement.org. https://www.scopus.com.

100

R. Miranda et al.

2. Publications that were not produced in the last 5 years (from 2018) or have not yet been fully published. 3. Publications that aren’t from the fields of Computer Science or Engineering. 4. Publications that aren’t Articles or Reviews/Surveys and aren’t written in English. 5. Publications that don’t focus on the variables studied or are out of context. 6. Publications that are not in accordance with the European Union’s General Data Protection Regulation (GDPR).

3

Results

The initial bibliographic research identified 14,840 studies, which were then subject to the defined eligibility criteria. The first five criteria were applied using Scopus’s filtering system, resulting in a subset of engineering and computer science articles or reviews that were freely accessible, written in English, and had relevant authors or citations in the last five years. Of the 575 studies, 526 were excluded due to not meeting the European Union’s standards for data extraction, manipulation, privacy, and data protection. The remaining 49 were fully read, with seventeen meeting the criteria of being related to crowdsensing and crowdsourcing applications in smart cities, namely, through geofences and locationbased notifications, and not developed in an enterprise context. Thus, 32 studies were excluded. Next, a summary of the results found in the review to answer the research questions will be presented. Ande et al. [3] made an overview and reference guide for IoT systems, particularly considering security issues. The authors show in this study that the Internet of Things is the result of improvements in computing, communication technologies, and the Internet, all which are combined with the human desire to better our quality of life. Amaxilatis et al. [2] present a solution for deploying and managing crowdsensing campaigns and experiments across cities federated in the OrganiCity facility. As a result, the authors concluded that the system allows developers to reverse course if they notice that their campaign design is underperforming through the system. Cheng et al. [5] propose a collaborative geofence site selection (CGSS) method for city management to designate geofence sites for dock-less shared bikes in the city. In this scenario, geofences are used to determine the best sites for renting bicycles and scooters to maximize user satisfaction with the service’s availability and coverage for the city’s expansion. P´ anek [19] describes the timeline of development and deployment of an emotional mapping approach, a methodology that allows individuals to initiate a map-based dialogue concerning the current and future state of public space, relying solely on their experiences. The author concluded that emotional maps, as a tool for crowdsourcing, provide an easy-to-use setting for social participation while fostering a sense of belonging to a specific social group or community.

Smart Cities Using Crowdsensing and Geoferenced Notifications

101

Fernandes et al. [6] developed a platform to be used via a mobile application that allows its users to obtain information regarding the current and future status of the city, as well as the provision of smart notifications with useful information taking into account the context of each user. In their paper, Foschini et al. [7] introduced ParticipAct, a mobile crowdsensing technology that uses edge nodes to calculate potentially hazardous crowd situations. This system requires users to have a sensing client application installed on their phone, and when sensing campaigns created by researchers or platform administrators are initiated, the collected data are sent to a centralized cloud server. Kousiouris et al. [12] leveraged social network data to find large crowd concentration (LCC) events. In this approach, the user data, coming from a Smart Transportation platform, is linked with social network activity peaks to identify large crowd concentration events that might affect the user journey. The authors concluded that this kind of mechanism can be used, for example, by the caregivers of a disabled user to alter the planned journey to avoid congestion, confusion and limited mobility circumstances. Roman et al. [20] suggested a mobile on-street parking spot identification system that is scalable and manageable when compared to a stationary sensor method. A supervised learning system that analyses the structure of the sonar trace to distinguish parked automobiles from road clutter, transmits this information to a central server that generates a map of parking occupation. The authors claim this mobile sensor system could be a crowdsourcing approach in the future, where members of the public can install sensors on their vehicles and collect data, with appropriate compensation in the form of reduced parking costs or simply free access to the parking apps. Picaut et al. [17] propose a crowdsourcing-based alternative method for assessing the noise environment. A smartphone application and a data infrastructure have been developed specifically for gathering physical and perceptual information about the sound environment. Each contributor’s data feeds a community database, which allows for a more detailed representation of the sound environment in space and over time than by using traditional numerical modelling methodologies. Kirimtat et al. [11] overviewed smart city projects by examining key concepts and data management strategies. The authors concluded that based on current developments in scientific studies, there is still a lack of scientific reports on smart floating cities, which seems to be good candidates for future smart cities. Pilloni et al. [18] presented the main aspects related to Industry 4.0. The authors explore the effects, advantages and challenges that IoT, crowdsourcing and crowdsensing, big data and other technologies will have on industrial processes. Yang et al. [25] present a perspective which displays crowdsensing as a technique in which a significant number of people using mobile devices with sensors share sensory data to measure, analyse, or infer any issue of common interest.

102

R. Miranda et al.

Shahrour and Xie [21] present and discuss the role of the IoT and crowdsourcing in constructing smart cities by emphasizing the role of data in smart city solutions, and the use of IoT and mobile crowdsourcing in smart city applications. The authors concluded that smart city development still needed more collaboration between smart city technology-centred research, which is primarily based on the IoT, and smart city citizen-centred research, which is primarily based on crowdsourcing. Wang et al. [24] focused on citizen recruiting through social network. Citizen engagement is vital to any crowdsensing project’s success, and it has become a key study area. In a study by Ismagilova et al. [10], the implications of Information Systems in smart cities are explored. The article covers a variety of topics, such as smart mobility, smart living, smart environment, smart citizens, smart government, and smart architecture, while also examining associated technologies and concepts. Shortcomings in existing research are revealed along with potential future developments. Nizetic et al. [14] conducted a comprehensive analysis of the Internet of Things (IoT) and its potential, difficulties, and worries in the context of a smart and sustainable future. Through their review, the authors improved the knowledge of existing technological advancements in IoT application areas, as well as the ecological ramifications of enlarging IoT product use. Ultimately, they ascertained that this study was beneficial in furthering understanding of the topic.

4

Discussion

This section explores and discusses the findings of the review against the research questions. The first subsection addresses the first research question and discusses the role and purpose of crowdsourcing, namely geofences and location-based notifications, in smart cities. The second presents the strategies and tools that have already been developed and used to implement and integrate mechanisms of crowdsourcing in smart cities. The last is related to how crowdsensing implementations can improve citizens’ lives in a smart city, and how to make citizens active contributors. 4.1

Role and Purpose of Crowdsensing and Crowdsourcing, Namely Geofences and Location-Based Notifications, in Smart Cities

According to Pilloni et al., MCS includes both mobile crowdsensing and mobile crowdsourcing [18]. The author states that “mobile crowdsensing makes use of mobile sensors to collect data from a crowd of different sources”, and that “in mobile crowdsourcing, users are required to complete a task, usually consisting of providing feedback”. MCS has been widely adopted in smart cities, personal health care, and environment monitor areas [9].

Smart Cities Using Crowdsensing and Geoferenced Notifications

103

Crowdsensing is a technique in which a large number of people using mobile devices with sensors share sensory data to measure, analyse, or infer any issue of common interest [25]. In fact, in recent years, mobile crowdsensing has emerged as one of the most popular paradigms for urban sensing, “allowing the collection and sharing of great amounts of data and the monitoring/detection of citizenship habits and movements in urban environments” [7]. In smart cities, crowdsourcing initiatives enable citizens’ active and passive engagement in data collection about the city’s functioning, services, quality of life, and environment. Mobile crowdsourcing relies on human sensing capacity, which combines sensors with human feedback and analysis. This is particularly valuable when users are assessing urban concerns such as the quality of urban services and the impact of decisions on users’ quality of life [21]. Shahrour and Xie [21] claim that crowdsourcing contributes to three key success factors of smart city projects. The first is related to data collection, as mobile crowdsourcing enables the development of cost-effective and high-quality monitoring systems for urban infrastructures, services, and the environment. Furthermore, user feedback based on sensor data is essential for capturing people’s demands and preferences, as well as determining the true impact of smart city projects on inhabitants’ quality of life. The second factor concerns citizens’ involvement in local development and activities, as local governments can get citizens’ ideas and feelings about smart city projects and their real-world impact using mobile crowdsourcing. The third factor involves the creation of smart applications based on crowdsourcing, like smart navigation, real-time public transportation, carpooling, risk warnings, emergency operations, and disturbance alerts. By developing crowdsourcing-based and cost-effective monitoring systems as an alternative to traditional smart city monitoring systems, mobile crowdsourcing could help speed up the deployment of smart city projects. According to Amaxilatis et al. [2], the use of IoT devices together with smartphone apps is a crucial enabler for the creation and consolidation of innovative ecosystems within cities, as it allows users to validate and embrace new services. The authors concluded that because people carry and utilise their smartphones daily, creating a significant potential for researchers to collect complex and intelligent observations of an urban environment. To summarize, the research suggests that mobile crowdsensing and mobile crowdsourcing have been widely adopted in smart cities, personal health care, and monitor areas. Crowdsensing allows for the collection and sharing of sensory data to measure, analyze, or infer any issue of interest. Crowdsourcing contributes to key success factors of smart city projects related to data collection and citizens’ involvement. The use of IoT devices with smartphone apps is also seen as a crucial enabler for the creation and consolidation of innovative ecosystems within cities.

104

4.2

R. Miranda et al.

Strategies and Tools that have Already Been Developed and Used to Implement and Integrate Mechanisms of Crowdsensing and Crowdsourcing in Smart Cities

Amaxilatis et al. [2] present a solution for deploying and managing crowdsensing campaigns and experiments across cities federated in the OrganiCity facility. The authors present a tool to help those planning crowdsensing campaigns define more exactly what they want as input from those taking part in such activities, as well as extra feedback that allows them to characterize their success numerically and qualitatively. Experimenters can see if participants’ contributions stick to the spatiotemporal limits they set, and participants can follow their contributions. As a result, the authors concluded that the system allows developers to reverse course if they notice that their campaign design is underperforming through the system. Cheng et al. [5] propose a collaborative geofence site selection (CGSS) method for city management to designate geofence sites for dock-less shared bikes in the city. It initially selects hotspots using a density-based and collaboration-inspired algorithm, and then assigns geofence sites to the highest-ranking hotspots. In this scenario, geofences are used to determine the best sites for renting bicycles and scooters to maximize user satisfaction with the service’s availability and coverage for the city’s expansion. Fernandes et al. [6] developed a platform to be used via an application mobile device that allows its users to obtain a set of information regarding the current status and future of the city, as well as the provision of smart notifications with useful information taking into account the context of each user. The platform also has a gamification mechanism that includes the subjective perception, evaluation and satisfaction of each user concerning the city in which they live, to promote the continued use of the platform in question. In 2021, Foschini et al. introduced ParticipAct, a mobile crowdsensing platform utilizing edge nodes to identify potential hazardous crowd scenarios [7]. This platform is especially helpful in emergency scenarios, such as the COVID19 pandemic, enabling people to avoid dangerous crowd situations, and suggesting people safer routes or places. In ParticipAct the users have a sensing client application installed on their smartphones and they send collected information to a centralized cloud server, based on targeted sensing campaigns created by researchers and platform administrators. In smart city applications, Kousiouris et al. [12] leveraged social network data to find large crowd concentration (LCC) events. In this approach, the user data, in the form of planned routes and locations, coming from a Smart Transportation platform, is linked with social network activity peaks to identify large crowd concentration events that might affect the user journey. For this project, the public bus transport authority has implemented Route Monitoring of individuals with special needs when they are taking buses throughout the city. If a large gathering of people is detected, the caregivers will be notified. Thus, the caregivers may alter the planned journey of the user to avoid congestion, confusion and limited mobility circumstances.

Smart Cities Using Crowdsensing and Geoferenced Notifications

105

Roman et al. [20] proposed a mobile on-street parking space detection system that is scalable and manageable when comparing its performance with a fixed sensing solution. Ultrasonic sensors are installed on the passenger side of a car to determine the distance between the vehicle and the next roadside object. A supervised learning system that analyses the structure of the sonar trace to distinguish parked automobiles from road clutter, transmits this information to a central server, which creates a parking occupancy map. This mobile sensor system could be a crowdsourcing approach in the future, where members of the public can install sensors on their vehicles and collect data, with appropriate compensation in the form of reduced parking costs or simply free access to the parking apps. Picaut et al. [17] propose a crowdsourcing-based alternative method for assessing the noise environment. To collect physical and perceptual data on the sound environment, a smartphone application and a spatial data infrastructure have been designed specifically for this purpose. Each contributor’s data feeds a community database, which should allow for a more detailed representation of the sound environment in space and over time than those obtained conventionally by using traditional numerical modelling methodologies. This research has identified various crowdsourcing methods in smart cities, including deploying and managing crowdsensing campaigns and experiments, collaborative geofence site selection for bike sharing, emotional mapping for public space dialogue, a mobile platform for obtaining city information and user satisfaction, edge-enabled mobile crowdsensing for dangerous crowd situations, leveraging social network data for large crowd concentration events, mobile on-street parking space detection, and crowdsourcing-based noise environment assessment. These methods utilize a combination of mobile devices, sensors, and community engagement to collect and analyse data for smart city applications. 4.3

Improvement of Citizens’ Lives in a Smart City Through Crowdsensing and Strategies to Make Citizens Active Contributors

At first, smart cities were simply cities with economic development objectives; later, the notion expanded to encompass the usage of information and communications technology (ICT) within city infrastructures, and finally, the concept became more citizen-centric. Today, smart cities incorporate technology into their infrastructures, allowing them to better serve their citizens by providing more efficient services to citizens, monitoring and optimizing existing infrastructure, increasing collaboration among various economic actors, and encouraging innovative business models in both the private and public sectors [3]. Mobile crowdsensing (MCS) has the potential to significantly improve citizens’ daily lives while also providing urban civilizations with new perspectives. Furthermore, the IoT paradigm is a suitable solution for a wide deployment of sensing infrastructure that enables smart city applications [4,14]. In smart cities, citizens can not only interact and engage with services but also provide data for these services via crowdsensing [10]. MCS makes use of human intelligence, which

106

R. Miranda et al.

has a better awareness of context than typical sensor networks. As a result, citizens’ active participation can improve the spatial coverage of existing deployed sensing systems without requiring further investments, according to Capponi et al. [4]. MCS has the potential to improve the living conditions of a smart city through numerous applications, such as emergency management and prevention, environmental monitoring, health care and wellbeing, e-commerce, indoor localization, intelligent transportation systems, mobile social networks, public safety, unmanned vehicles, urban planning, waste management, and Wi-Fi characterization. Capponi et al. [4] have identified these domains as the most promising for the use of MCS. The success of any crowdsensing implementation depends on citizen engagement and contribution [4,7,24]. Gathering more data leads to more thorough outcomes and, consequently, better quality data sets [7]. In this way, citizen recruitment and data contribution incentives are critical to the success of MCS programs and how to attract a big group of citizens as active contributors has become a major research topic [4,24]. A common strategy for increasing and reinforcing participation is through a gamification technique. In this way, the adoption of reward programs encourages and stimulates the users to complete the task, augmenting the quantity and the quality of the gathered information [7]. Many strategies have been proposed to stimulate participation and make citizens active contributors, and they can be classified into three categories, according to Capponi et al. [4]: entertainment, service, and money. The first category considers methods that stimulate participation by turning sensing tasks into games. The service category consists of providing services to the users in exchange for their data. Finally, monetary incentives reward users for their contributions with money. The authors concluded that these strategies are highly dependent on the type of application and that, in general, rewards should be based on the quality of contributed data and determined by the demand-supply relationship. Hu et al. [9] created a three-stage Stackelberg game to keep the number of participants constant by viewing the MCS situation as a sensory data market. Players in the three-stage Stackelberg game are divided into two groups: monthly-pay participants and instant-pay participants. This ensures that participants who pay monthly will continue to contribute with sensory data. Furthermore, the game protects the sensory data market’s fairness by utilizing a safe reward distribution mechanism supported by blockchain technology. To conclude, smart cities use information and communication technologies to offer efficient services to citizens, monitor and optimize infrastructure, promote collaboration, and encourage innovation. Mobile crowdsensing provides an effective means for gathering data in smart cities, with citizens playing an active role. Crowdsensing solutions can enhance emergency management, environmental monitoring, health care and wellbeing, e-commerce, and other domains within smart cities. Citizen engagement is critical to its success, and gamification tech-

Smart Cities Using Crowdsensing and Geoferenced Notifications

107

niques, such as reward programs, can promote participation. Strategies to promote participation can be classified into the categories of entertainment, service, and monetary incentives, with rewards based on the quality of contributed data.

5

Conclusion

Smart cities are considered a crucial element in achieving the sustainability of cities and urban areas with high-population densities. Thus, the involvement of residents in the transition to smarter cities is essential and one of the main mechanisms for this is the application of crowdsensing and crowdsourcing strategies. This article details a systematic review that aims to analyze the literature on smart cities and their implementation using crowdsensing and crowdsourcing. The review followed the PRISMA model and used the Scopus database as the primary source of information. Out of the 14,840 studies initially identified, 17 were finally selected based on the previously defined research questions after applying the PRISMA methodology. The aim of the research questions was to evaluate the current state of crowdsensing and crowdsourcing in smart cities. Based on the studies collected, it was found that crowdsensing enables the collection and exchange of massive amounts of data, which can be utilized to track citizens’ behavior and movement in urban areas. Crowdsourcing initiatives enable citizens’ active and passive involvement in gathering data about the quality of life and services in the city, ensuring that they function at their best. Furthermore, several strategies and tools that have already been developed and used to implement and integrate mechanisms of crowdsensing and crowdsourcing in smart cities, were found in the literature. Finally, in most of the analyzed studies, all the authors agreed that the implementation of crowdsensing initiatives can improve the citizens’ lives in a smart city. It was also identified the most promising application domains where MCS can play an important role, for example, environmental monitoring, e-commerce, health care and well-being, indoor localisation, intelligent transportation systems, public safety, urban planning, and waste management. This project intends to examine how crowdsensing, crowdsourcing, and geofences can be incorporated into the infrastructure and framework of smart cities. Using the compiled information, the next step is to create a platform that integrates these concepts to form a valuable information system, enhancing the smart city concept and improving the quality of life for its residents. Acknowledgements. This work has been supported by FCT-Funda¸ca ˜o para a Ciˆencia e Tecnologia within the R&D Units Project Scope: UIDB/00319/2020 and the project “Integrated and Innovative Solutions for the well-being of people in complex urban centers” within the Project Scope NORTE-01-0145-FEDER-000086. We would like to thank also the Guimar˜ aes city hall for making available multiple datasets.

108

R. Miranda et al.

References 1. Alves, C., Chaves, A., Rodrigues, C., Ribeiro, E., Silva, A., Dur˜ aes, D., Machado, J., Novais, P.: Survey for big data platforms and resources management for smart cities. In: Bringas, P.G., Garc´ıa, H.P., de Pis´ on, F.J.M., Flecha, J.R.V., Lora, ´ Mart´ınez-Alvarez, ´ A.T., de la Cal, E.A., Herrero, A., F., Psaila, G., Quinti´ an, H., Corchado, E. (eds.) Hybrid Artificial Intelligent Systems—17th International Conference, HAIS 2022, Salamanca, Spain, Proceedings. Lecture Notes in Computer Science, vol. 13469, pp. 393–404. Springer (2022). https://doi.org/10.1007/978-3031-15471-3 34 2. Amaxilatis, D., Mylonas, G., Diez, L., Theodoridis, E., Guti´errez, V., Mu˜ noz, L.: Managing pervasive sensing campaigns via an experimentation-as-a-service platform for smart cities. Sensors (Switzerland) 18 (2018). https://doi.org/10.3390/ s18072125 3. Ande, R., Adebisi, B., Hammoudeh, M., Saleem, J.: Internet of things: evolution and technologies from a security perspective. Sustain. Cities Soc. 54 (2020). https://doi.org/10.1016/j.scs.2019.101728 4. Capponi, A., Fiandrino, C., Kantarci, B., Foschini, L., Kliazovich, D., Bouvry, P.: A survey on mobile crowdsensing systems: challenges, solutions, and opportunities. IEEE Commun. Surv. Tutor. 21, 2419–2465 (2019). https://doi.org/10.1109/ COMST.2019.2914030 5. Cheng, G., Guo, Y., Chen, Y., Qin, Y.: Designating city-wide collaborative geofence sites for renting and returning dock-less shared bikes. IEEE Access 7, 35596–35605 (2019). https://doi.org/10.1109/ACCESS.2019.2903521 6. Fernandes, B., Neves, J., Analide, C.: Safecity: a platform for safer and smarter cities. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 12092 LNAI, pp. 412–416 (2020). https://doi.org/10.1007/978-3-030-49778-1 37 7. Foschini, L., Martuscelli, G., Montanari, R., Solimando, M.: Edge-enabled mobile crowdsensing to support effective rewarding for data collection in pandemic events. J. Grid Comput. 19 (2021). https://doi.org/10.1007/s10723-021-09569-9 8. Google: geofencing API.: (2022). https://developers.google.com/location-context/ geofencing 9. Hu, J., Yang, K., Wang, K., Zhang, K.: A blockchain-based reward mechanism for mobile crowdsensing. IEEE Trans. Comput. Soc. Syst. 7, 178–191 (2020). https:// doi.org/10.1109/TCSS.2019.2956629 10. Ismagilova, E., Hughes, L., Dwivedi, Y.K., Raman, K.R.: Smart cities: advances in research - an information systems perspective. Int. J. Inf. Manag. 47, 88–100 (2019) 11. Kirimtat, A., Krejcar, O., Kertesz, A., Tasgetiren, M.F.: Future trends and current state of smart city concepts: a survey. IEEE Access 8, 86448–86467 (2020). https:// doi.org/10.1109/ACCESS.2020.2992441 12. Kousiouris, G., Akbar, A., Sancho, J., Ta-shma, P., Psychas, A., Kyriazis, D., Varvarigou, T.: An integrated information lifecycle management framework for exploiting social network data to identify dynamic large crowd concentration events in smart cities applications. Fut. Gener. Comput. Syst. 78, 516–530 (2018). https:// doi.org/10.1016/j.future.2017.07.026

Smart Cities Using Crowdsensing and Geoferenced Notifications

109

13. Miranda, R., Ramos, V., Ribeiro, E., Rodrigues, C., Silva, A., Dur˜aes, D., Analide, C., Abelha, A., Machado, J.: Crowdsensing on smart cities: a systematic review. In: Advances in Artificial Intelligence—IBERAMIA 2022: 17th Ibero-American Conference on AI, Cartagena de Indias, Colombia, Proceedings, pp. 103–106. Springer (2023) ˇ c, P., de-Ipi˜ 14. Niˇzeti´c, S., Soli´ na Gonz´ alez-de Artaza, D.L., Patrono, L.: Internet of things (IoT): opportunities, issues and challenges towards a smart and sustainable future. J. Clean. Prod. 274 (2020). https://doi.org/10.1016/j.jclepro.2020.122877 15. Page, M.J., McKenzie, J.E., Bossuyt, P.M., Boutron, I., Hoffmann, T.C., Mulrow, C.D., Shamseer, L., Tetzlaff, J.M., Akl, E.A., Brennan, S.E., Chou, R., Glanville, J., Grimshaw, J.M., Hr´ objartsson, A., Lalu, M.M., Li, T., Loder, E.W., MayoWilson, E., McDonald, S., McGuinness, L.A., Stewart, L.A., Thomas, J., Tricco, A.C., Welch, V.A., Whiting, P., Moher, D.: The prisma 2020 statement: an updated guideline for reporting systematic reviews. BMJ 372 (2021). https://doi.org/10. 1136/bmj.n71 16. Pereira, P., Linhares Silva, A., Machado, R., Silva, J., Dur˜ aes, D., Machado, J., Novais, P., Monteiro, J., Melo-Pinto, P., Fernandes, D.: Comparison of different deployment approaches of FPGA-based hardware accelerator for 3d object detection models. In: Progress in Artificial Intelligence: 21st EPIA Conference on Artificial Intelligence, EPIA 2022, Lisbon, Portugal, Proceedings, pp. 285–296. Springer (2022) 17. Picaut, J., Fortin, N., Bocher, E., Petit, G., Aumond, P., Guillaume, G.: An openscience crowdsourcing approach for producing community noise maps using smartphones. Build. Environ. 148, 20–33 (2019). https://doi.org/10.1016/j.buildenv. 2018.10.049 18. Pilloni, V.: How data will transform industrial processes: crowdsensing, crowdsourcing and big data as pillars of industry 4.0. Fut. Internet 10 (2018). https:// doi.org/10.3390/fi10030024 19. P´ anek, J.: Emotional maps: participatory crowdsourcing of citizens’ perceptions of their urban environment. Cartogr. Perspect. 2018, 17–29 (2019). https://doi.org/ 10.14714/CP91.1419 20. Roman, C., Liao, R., Ball, P., Ou, S., Heaver, M.D.: Detecting on-street parking spaces in smart cities: performance evaluation of fixed and mobile sensing systems. IEEE Trans. Intell. Transp. Syst. 19, 2234–2245 (2018). https://doi.org/10.1109/ TITS.2018.2804169 21. Shahrour, I., Xie, X.: Role of internet of things (IoT) and crowdsourcing in smart city projects. Smart Cities 4, 1276–1292 (2021). https://doi.org/10.3390/ smartcities4040068 22. Silva, G.O., Rocha, A.M.A., Witeck, G.R., Silva, A., Dur˜ aes, D., Machado, J.: On tuning the particle swarm optimization for solving the traffic light problem. In: Computational Science and its Applications—ICCSA 2022 Workshops: Malaga, Proceedings, Part II, pp. 68–80. Springer, Spain (2022) 23. Sousa, R., Lopes, D., Silva, A., Dur˜ aes, D., Peixoto, H., Machado, J., Novais, P.: Sustainable and social energy on smart cities: systematic review. In: Guarda, T., Portela, F., Augusto, M.F. (eds.) Advanced Research in Technologies, Information, Innovation and Sustainability—Second International Conference, ARTIIS 2022, Santiago de Compostela, Spain, Revised Selected Papers, Part II. Communications in Computer and Information Science, vol. 1676, pp. 72–84. Springer (2022). https://doi.org/10.1007/978-3-031-20316-9 6, https://doi.org/10.1007/978-3-03120316-9 6

110

R. Miranda et al.

24. Wang, J., Wang, F., Wang, Y., Zhang, D., Wang, L., Qiu, Z.: Social-networkassisted worker recruitment in mobile crowd sensing. IEEE Trans. Mob. Comput. 18(7), 1661–1673 (2019). https://doi.org/10.1109/TMC.2018.2865355 25. Yang, M., Zhu, T., Liang, K., Zhou, W., Deng, R.H.: A blockchain-based location privacy-preserving crowdsensing system. Fut. Gener. Comput. Syst. 94, 408–418 (2019). https://doi.org/10.1016/j.future.2018.11.046

Understanding Urban Mobility Habits and Their Influencing Factors on a University Campus in Argentina Gabriela Pesce1(B) , Florencia Pedroni1,2 , María Andrea Rivero1 , Héctor G. Chiacchiarini3,4 , Yamila S. Grassi5 , and Mónica F. Díaz5,6 1 Departamento de Ciencias de la Administración e Instituto de Investigación en Ciencias de la

Administración (IICA, UNS), Universidad Nacional del Sur (UNS), San Andrés 800, 8000 Bahía Blanca, Buenos Aires, Argentina [email protected] 2 CONICET, San Andrés 800, 8000 Bahía Blanca, Buenos Aires, Argentina 3 Departamento de Ingeniería Eléctrica y de Computadoras, Universidad Nacional del Sur, San Andrés 800, 8000 Bahía Blanca, Buenos Aires, Argentina 4 Instituto de Investigaciones en Ingeniería Eléctrica “Alfredo C. Desages”, Universidad Nacional del Sur, and CONICET, San Andrés 800, 8000 Bahía Blanca, Buenos Aires, Argentina 5 Planta Piloto de Ingeniería Química (PLAPIQUI), Universidad Nacional del Sur, and CONICET, C. La Carrindanga km 7, 8000 Bahía Blanca, Buenos Aires, Argentina 6 Departamento de Ingeniería Química, Universidad Nacional del Sur (UNS), Av. Alem 1253, 8000 Bahía Blanca, Buenos Aires, Argentina

Abstract. In recent years, various social and environmental phenomena have highlighted the relevance of mobility impacts in cities. This paper aims to analyze the urban mobility decisions of a university community and its primary influencing factors from a sustainable development perspective. We conducted a case study of the academic community belonging to Universidad Nacional del Sur in the city of Bahía Blanca, Argentina. Based on multiple triangulated primary and secondary data sources, we diagnosed how the university community moves. Our findings show that private combustion vehicles (41–50%), public transport by bus (29–31%), and active mobility (20–26%) are the most commonly used means of transport, with electric mobility accounting for less than 2%. About influencing factors in the decision of the type of mobility (active, public, or private), we identified several determinants, including age, gender, trip distance, ownership of a car or bicycle, and factors perceived as most relevant when deciding on the means of transport, such as travel time, weather conditions, comfort, and the awareness of the environmental impact of mobility. This article contributes to understanding how urban mobility can be improved to promote sustainable development. Keywords: Sustainable urban mobility · Daily transport · Modal choice determinants · Medium-sized city · Developing country

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 111–123, 2023. https://doi.org/10.1007/978-3-031-36957-5_10

112

G. Pesce et al.

1 Introduction Nowadays society is becoming concern about pollution, and daily traffic problems in cities are driving the development of new trends in mobility. There is an increase in the use of various modes of transport in urban areas, including electric and smaller vehicles, such as motorcycles, bikes, and skateboards, as well as shared mobility. These modes of transport have different environmental effects, with motorised modes contributing the most to CO2 emissions and climate change [1]. In this sense, the analysis of mobility decisions and modal distribution acquires its maximum relevance since it not only shows what means of transport are being used but also which population groups make the trips and what reasons they have for them [1]. Although urban mobility trends have been evaluated from the point of view of their convenience at the aggregate level and for large cities [2, 3], their study in small and medium-sized cities is also relevant from the perspective of sustainable development, a complex concept due to its multidimensionality including three pillars: ecological, economic, and socio-political-cultural. A simple analysis of these dimensions does not reveal a dominant means of transport, being necessary a multidisciplinary analysis to understand how improve urban mobility to contribute to sustainable development. A first approach to this problem can be made from the study of urban mobility on university campuses [4, 5]. These spaces are considered models of micro-cities that present within them the same phenomena as the cities in which they are located, but on a smaller scale [6]. It is known that each city, depending on its particularities (size, urbanization, economy, etc.), will present its own mobility problems. In this sense, universities play a key role in promoting sustainable mobility in the cities where they are located since, it has the ability to encourage members of the university community to be aware that their actions and attitudes have an impact on the environment [7]. Within the framework of a broader research project1 , this paper aims to analyze the urban mobility decisions of a university community and its main influencing factors, from the perspective of sustainable development. To fulfill the objective, we work with a case study of the community attending a university campus belonging to the Universidad Nacional del Sur (UNS), located in the city of Bahía Blanca, Argentina. Results are obtained from processing primary and secondary data from multiple sources. The work recognizes theoretical and empirical contributions. At the theoretical level, it systematizes knowledge on urban mobility, compiling studies about modal choice and, in particular, daily mobility on university campuses in developed and emerging countries. As empirical implications, the description of the current mobility of the studied university campus, as well as the factors influencing mobility decisions, provides relevant information to policymakers for designing incentives to reduce the impact of mobility decisions, thus contributing to the fulfillment of various Sustainable Development Goals (SDGs): sustainable cities (SDG11), climate change (SDG13), energy (SDG7), and responsible production and consumption (SDG12). 1 This work is part of an inter-institutional applied research project approved by Inter-American

Organization for Higher Education, in a specific call for university collaborative research on the application of the sustainable development goals of United Nations. The objective of the project is to multidimensionally analyze sustainable urban mobility proposals and collaborate with their implementation in Bahía Blanca (Argentina).

Understanding Urban Mobility Habits and Their Influencing Factors

113

The article presents in Sect. 2 the conceptual framework, in Sect. 3 the methodological details of the research, in Sect. 4 the results organized as follows: (4.1) the characteristics of the study case, (4.2) a descriptive study on the urban mobility of the university community, and (4.3) an analysis of the factors that influence urban mobility decisions. Finally, conclusions and future developments are mentioned.

2 Conceptual Framework The modal choice is a decision process to choose between different transport alternatives, which is determined by a combination of factors that comes from three major approaches. The rationalist approach in modal choice assumes that travelers take decisions based on utility maximization attained by minimizing travel time and costs. The socio-geographical approach explicitly introduces a spatial component into the modal choice decision process and starts from the activity schedule of individuals or households. The socio-psychological approach considers attitudes, intentions, and habits of individuals with regard to the available means of transport as relevant factors. In sum, the modal choice is determined by a combination of interrelated factors: individual socio-demographic issues, spatial characteristics, and socio-psychological features [8]. In their literature review, [8] identify a list of 26 determinants of modal choice and organize them according to the times each factor has been studied and were found significant. It appears that objective and straightforward determinants, like socio-demographic features (age, gender, employment, income, etc.) and car availability, are more frequently studied compared to subjective ones (familiarity, trip chaining, lifestyle, etc.). Empirical studies have analyzed factors affecting modal choice in urban mobility with different variables of interest. In Kalamaria (Greece), [9] using a binary explanatory variable (bus = 1, car = 0) find a general preference of people for cars over public transport, and the main factor affecting the preference of respondents toward passenger cars is the availability of parking space. Moreover, gender, age, and trip purpose are found significant determinants with a higher preference for the car for male respondents, for 35–44 years’ people, and for the general population in the case of work trips. In Barcelona (Spain), [10] using a proximity trip variable (walk < 10 min) also find that gender, age, and trip purpose are significant factors of daily mobility, added to the access to private transport. In this sense, the use of proximity trips increases for older people, women, lower income groups, and for personal journeys. In an ampler study of 98 cities in India using continue and binary variables (out-of-pocket travel expenditure, motorized = 1, private = 1), [11] find that income is the most important determinant of the amount of trips and the use of motorized and private transport. Furthermore, empirical literature also examines the daily mobility on university campuses (Table 1), mostly in a descriptive way. However, some articles also identify its determinants factors through correlation or multivariate analysis ([*] in Table 1). Maciejewska et al. [1] find significant differences in university community modal choice by role and gender. Students show higher use of public transport than staff. Women use less active transport and more public transport than men, in fact, they are the largest users of public ones. Zhou [12] also report that gender and status (undergraduate vs. graduate) are associated with both commuting distance and alternative transport usage.

114

G. Pesce et al.

In particular, female or graduate students tend to be less likely to use alternative transport and undergraduate students tend to have a shorter commute and use more alternative transport. According to [13], male students are more likely to bike to campus while females are generally more likely to drive. Distance also increases the probability of driving instead of walking, and parking was the dominant positive variable associated with the probability of driving to campus. Aside from gender, distance, and parking, differences in modal choice decisions are also related to different kinds of students. Graduates students are less likely to use their car to commute to campus, and more likely to walk or bike. Moreover, in students with children, the number of kids has a positive correlation with the probability of driving and a negative correlation with walking and biking. Finally, some articles identify determinant factors of university modal choice, without descriptive results. In Universidad Autónoma de Barcelona (UAB), [21] find that age and weekly attendance to the campus were key factors in the use of cars in daily mobility. Moreover, statistical correlations were seen between car-weekly-km travelled by the university community and factors such as: urban density (+), distance to the campus (+), gender-female- (+), age-older than 55 years- (+), role (staff travels more km than students), car availability (+), weekly attendance at the campus—5 days or more(+), daily stay—8 h per day or more- (+). According to [22] estimation with data from the University of Deusto, the use of public transport is negatively affected by the minutes between services, trip time, and ticket price.

3 Methodology A quantitative research is proposed, through a case study on the university campus Altos de Palihue belonging to UNS [23]. We collect primary and secondary data. The diagnosis with secondary information compresses survey activities for the entire city of Bahía Blanca, through the analysis of meteorological and public transport data. For meteorological aspects, data on rains and winds in the city for the last 10 years (2012– 2021) are analyzed. To diagnose transport, data provided by the Municipality of Bahía Blanca (MBB) on the use of public transport for the years 2019 and 2022 (pre- and post-pandemic) are considered. The analysis of the number of trips is carried out in general for all passengers and discriminating also for university passengers (users of two particular lines: 503 and 519A and differential fee for students). Diagnosis continues with the collection and analysis of primary data on mobility to and from the Altos de Palihue university campus, from three instruments: direct observations, survey, and interviews. Triangulation of sources is sought to increase the constructs’ validity, by providing various evaluations of the mobility phenomenon [24]. Direct observations and a protocolized count of incoming and outgoing passengers segregated by means of transport are carried out at the two entrances to Altos de Palihue university campus (by San Andrés Street and Cabrera Street) through the filming of five daytime time slots (7:30; 11:30; 13:30; 15:30; 17:30). The recordings were in September 2022 during the the second university semester. It is complemented by night-time footage from the MBB security camera on Payró and San Andrés streets. An online survey is prepared and applied to the university community with around 25 questions divided into 4 sections: sociodemographic data, current mobility decisions;

Understanding Urban Mobility Habits and Their Influencing Factors

115

Table 1. Empirical background of mobility on university campuses. Art

Country/Univ

Transport

[12] [*]

USA, Univ. of California Private transport, public [students] transport, non-motorized transport, carpooling

Gender, role, student status, housing, trip time, motive, parking

[13] [*]

USA, University of Idaho [students]

Walking, car, bike

Gender, trip time, student kind, marital status, kids, distance, parking, car ownership

[1] [*]

Spain, Univ. Autónoma Barcelona

Public transport, private transport, walking, bike

Age, gender, motive, car availability, driver’s license

[7]

Spain, Univ. Rovira I Virgili [students]

Car, bus, train, walking

Gender, age, location, distance, motive, student area

[14]

Spain, Univ. Valladolid

Walking, public transport, car, bicycle, motorbike

Motive, role

[15]

Spain, Univ. Autónoma Barcelona

Public and private transport, and non-motorized transport

Trip time, preferences, barriers

[16]

Mexico, Univ. Valle de México [students]

Car, public transport, walking, bicycle

Location

[6]

Colombia, Univ. de Antioquia

Bus, walking, metro, car, motorbike, taxi, bike

Role, origin, destination, parking conditions

[17]

Ecuador, Univ. Técnica de Manabí

Public transport, private transport, bike, others

Schedule, parking conditions, problems

[18]

Ecuador, Univ. Nacional Public transport, walking, Distance, trip time, de Chimborazo multimodal, car, taxi, bike location, daily cost

[19]

Panama, Univ. de Panamá

[20]

Cuba, Univ. Tecnológica Public transport, microbus, Location, trip time, motive, de La Habana motorbike, car, walking daily cost

Multimodal, metrobus, car, bus

Factors

Location, role, trip time, extra activities

Notes: If not specified, data collection instruments used are surveys of the entire university community (students, professors, and administrative staff). Transport indicates the most used transport in descending order. Factors are the main elements covered in the article in a descriptive way, motive refers to the reasons for choosing certain means of transport. [*] Includes correlational or multivariate analysis of determinants. Source: own elaboration

propensity for other mobility alternatives, and problems identified in urban mobility and improvement proposals. This instrument is disclosed through various digital media between June and December 2022. We obtain 743 observations, which represent a statistically significant random sample with 97.5% confidence and 3.52% error, on an estimated university population of 18,507 people (15,412 undergraduate students).

116

G. Pesce et al.

Complementary, to complete the analysis of quantitative data on public transport and also to define questions of direct observation, we conduct interviews with five bus drivers that go to the university campus -two from 503 line and three of the free UNS bus(open questions about: times, days and months of greatest bus demand and variations in demand due to weather, bus stops with the highest number of passengers) and the head of the UNS Welfare Secretariat (queries about UNS mobility policies, transportation scholarships, planning and future development of the campus). Regarding ethical during data collection, the survey included an initial paragraph that guarantees anonymity and confidentiality of information to the respondents. The interviewees were previously asked for consent and they were informed about these privacy issues. For the recordings, we obtained the permission of the UNS authorities, and the recording persons were identified with credentials about the project. Data analysis is developed through a quantitative-econometric approach, including descriptive statistics (Sects. 4.1 and 4.2), correlation, and multivariate analysis (4.3). The last one is developed by running three probit regression models [M1, M2, M3], where the binary dependent variables show the propensity to use different means of transport. Dependent dummy variables take unit value when the individual makes, on average, at least 5 out of 10 trips with ACTIVE MOBILITY (M1), moving on foot or on pedal bicycle, PUBLIC MOBILITY (M2), by city line bus or university bus, and PRIVATE MOBILITY (M3), using combustion car, pick-up truck, cab, or taxi. The following variables are considered as independent mobility determinants and control variables: individual’s age (continuous), self-perceived gender (binary, 1 = female), role in which he/she attends campus (binary, 1 = graduate student), trip distance to campus (ordinal, with categories according to Fig. 1), economic level (ordinal, with three categories), bicycle ownership (binary), and car ownership (binary), and the factors (F) that are considered a priority when deciding which means of transport to use. All the variables associated with factors are binary with unit value when the individual chooses 1st or 2nd priority to each of the following: travel time, weather conditions, time flexibility, comfort, economic factor, risk of theft, risk of accident, and environmental impact of mobility. For explanatory variables, we obtain marginal effects which exhibit how (sign +/−) and how much each determinant influences the probability of each type of mobility. Stata software version 14 is used for data processing.

4 Results and Discussion 4.1 Description of the Case Study Altos de Palihue campus of UNS agglomerates the classroom activity of eight academic units and, in addition to the infrastructure of six academic units, it has four classroom buildings, a dining room, and a reading room, among other common facilities. The campus is placed in Altos de Palihue neighborhood in Bahia Blanca city. Bahía Blanca is a medium-sized city located in the southwest of Buenos Aires province, in Argentina, with an urban conglomerate that covers an area of 2,247 km2 and has 335,190 inhabitants [25]. Currently, it has no urban transport by train or formal shared mobility systems, except for buses.

Understanding Urban Mobility Habits and Their Influencing Factors

117

According to the data collected, the community that attends the university campus in different roles, considering the same person can attend in more than one role, is represented by 70% of undergraduate students, 25.6% of teachers, and in minority participations, researchers (9.7%), graduate students and/or scholarship holders of postgraduate careers (6.5%), administrative and support staff (4.6%). Regarding gender, 62.2% perceived themselves as female and 37.2% as male. In relation to the socioeconomic situation, it is substantially better than the macroeconomic aggregates of the country: 53.8% declare that they have sufficient economic resources for their basic needs, with a surplus for other expenses and/or savings, 41.6% only to cover their basic needs and the rest (4.6%) consider those insufficient for this. 4.2 Descriptive Analysis of Urban Mobility of the University Community Regarding the current university community’s mobility decisions, data on the frequency, time range, average distance, factors influencing mobility decisions, availability of means of transport, and their usage for commuting to the Altos de Palihue campus were obtained from various sources mentioned in the methodological section. In terms of the average distance covered, 65.8% of the participants travel a distance of less than 5 km to reach the campus, whereas 90.3% of them commuted for less than 10 km. People who had to travel longer distances are in the minority (see Fig. 1). 100%

20% 15% 10%

Probability (left axis) Cumulative probability (right axis)

80% 60% 40% 20%

0%

0% up to 1 km. up to 2 km. up to 3 km. up to 4 km. up to 5 km. up to 6 km. up to 7 km. up to 8 km. up to 9 km. up to 10 km. up to 11 km. up to 12 km. up to 13 km. up to 14 km. up to 15 km. up to 16 km. up to 17 km. up to 18 km. up to 19 km. up to 20 km. 20 to 30 km. 30 to 40 km. 40 to 50 km. + 50 km.

5%

Fig. 1. Histogram of average distances. Source: own elaboration

Regarding the frequency of attendance at the university campus, the survey data reveals that 51.7% of the participants visit the campus between 4 and 5 times a week. 27.1% of them attend 2 to 3 times a week, while 13.1% of them visit more than 5 times, which may include weekends and/or more than one visit per day. Only a small percentage of 8.2% visit the campus once a day. About the time range, the data on public transport provided by MBB shows a significant decrease in the use of public transport during weekends and holidays in the year before and after the pandemic (2019 and 2022). In terms of specific time periods, peaks in public transport usage are observed between 7–8 a.m., 1–2 p.m., and 6–7 p.m. for all passengers, and between 10–11 p.m. for university passengers. It is possible to

118

G. Pesce et al.

associate these peak periods with the times of entrance/exit for school and work activities (7–8 a.m., 1–2 p.m., 6–7 p.m.) and the end of the last university class (10–11 p.m.). The UNS academic community has great variability in the times of entry to the campus and the same person may report entering at different time ranges, which may be due to their attendance not adhering to a fixed schedule. The most frequently reported entry time range is between 8 and 9 a.m. (12.4% of entries), followed by 7–8 a.m. (10.8%), 6–7 a.m. (8.8%), and 1–2 p.m. (8.5%). These survey primary data are consistent with public transport data. Regarding departure times, nighttime schedules are more prevalent, with the most frequent being between 9 and 10 p.m. (13.3% of declared departures), 8–9 p.m. (10.9%), and 6–7 p.m. (10.7%). According to the means of transport currently used to travel to the campus, we triangulated two sources of information. First, survey responses—that provide an idea of individuals’ declarations about their typical mobility habits, considering the proportion of respondents who report regarding 10 average trips. Second, direct observation counting—that compute the entry and exit of people according to the mode of transport, considering a weighted average of the two campus access where recordings were made. In this case, the data is revealed rather than declared. The results presented in Table 2 are consistent between both data sources and indicate that the most commonly used modes of transport are private combustion vehicles, accounting for between 41% and 50%, public transport (buses) between 29% and 31%, and active mobility between 20% and 26%. Only less than 2% of individuals reported using electric mobility. Table 2. Mobility decisions by means of transport. Source: Own elaboration Means of transport

Declared flow (%)

Revealed flow (%)

Combustion vehicles

41.1

50.2

Private car

31.0

41.0

Car sharing

5.3



Pick-up truck

1.4

5.2

Taxi/Cab

1.8

1.2

Combustion motorcycle

1.5

2.8

Active mobility

25.7

20.2

On foot

19.4

16.7

Pedal bicycle

6.4

3.5

Public transport

31.3

28.9

Line bus

20.1



University bus

11.2



Electric mobility

1.9

0.7

Electric bicycle

0.4

0.0

Electric skateboard

0.9

0.4

Electric motorcycle

0.6

0.3

Understanding Urban Mobility Habits and Their Influencing Factors

119

Regarding the availability of means of transport, 45.6% of the participants stated that they have a private car and 35.8% have a pedal bicycle. Notably, 36.1% of the university community does not have any means of transport, while 22.7% of those surveyed stated that they have more than one means of transport available. The survey also sought to gauge the willingness of respondents to use other mobility alternatives, revealing that 79% are in favor of carpooling, and 47% would be willing to use public bicycles. The most important factors influencing the current mobility decisions, based on a normalized weighted index (NWI) according to the level of importance assigned, are travel time (NWI = 1.00), followed by weather conditions (NWI: 0.92), flexible schedule (NWI: 0.67), comfort (NWI: 0.56), and economic factors (NWI: 0.46). In particular, regarding the impact of weather conditions on the use of public bus transport, historical data obtained from both SMN and MBB suggest that the usage of buses tends to increase on days with higher wind speeds, with an 88% confidence level. 4.3 Analysis of Factor Influencing Urban Mobility Decisions Based on the Pearson correlation coefficient analysis conducted on the variables of interest (Table 3), it can be concluded that the interrelationships are mostly consistent with those reported in the literature review. To enhance the results, probit multivariate analyses (Table 4) have been conducted to explain the propensity to move actively (M1), to use public mobility (M2), or private combustion mobility (M3). The likelihood of ACTIVE MOBILITY is negatively influenced by age, female gender (in line with the results of [1, 13]), trip distance, car ownership, and recognition of comfort and time flexibility as key factors for mobility. On the other hand, the likelihood of walking or cycling is positively affected for those who own a bicycle and those who prioritize weather conditions and the environmental impact of mobility in their decisionmaking. In terms of marginal effects size, car ownership reduces the probability of active mobility by 29.10%, the female gender reduces it by 12.23%, while awareness of the environmental impact of mobility increases it by 25.16%. The probability of using PUBLIC MOBILITY, such as line bus or university bus, is negatively influenced by age, car ownership (−33.10%), and considering comfort as a key decision factor (−11.77%). In contrast, the probability of taking the bus increases with female gender status (+10.96%, as found in [1]), being an undergraduate student (+20.74%, also as found in [1]), and the distance to travel to reach the campus. The probability of using PRIVATE MOBILITY by combustion car is positively influenced by age (in line with findings from [9, 21]), economic level (as found in [11]), car ownership (+55.20%, like [21]), and prioritization of travel time and risk of theft in mobility decisions. However, the probability is negatively affected by prioritization of the economic factor and environmental impact (−19.74%) in mobility decisions.

120

G. Pesce et al. Table 3. Correlation matrix. Source: own elaboration.

Active mobility Public mobility –bus Private mobility –car Age Gender Student role Distance Economic level Bike ownership Car ownership F: Travel time F: Weather conditions F: Flexible schedule F: Comfort F: Economic factor F: Risk of theft F: Risk of accident F: Environment

Active mob.

Public Private mob. mob.

1.00 -0.10 -0.35 -0.19 -0.12 0.14 -0.17 -0.06 0.06 -0.32 -0.15 0.23 -0.04 -0.12 -0.01 0.04 0.01 0.08

1.00 -0.46 -0.37 0.17 0.42 0.15 -0.25 -0.16 -0.50 -0.06 0.08 -0.01 -0.22 0.22 -0.01 0.02 0.01

1.00 0.43 -0.07 -0.40 -0.01 0.30 0.12 0.69 0.21 -0.24 0.06 0.19 -0.19 0.01 0.01 -0.05

Age

1.00 -0.05 -0.77 0.00 0.16 0.11 0.45 0.20 -0.22 0.07 0.16 -0.08 -0.08 0.00 0.03

Gender

Student Economic Bike Car Distance role level ownership ownership The greater colour intensity, the greater the positive (red) or negative (blue) correlation

1.00 0.08 0.02 -0.11 -0.20 -0.12 0.03 0.00 0.03 -0.06 -0.04 0.04 0.05 -0.09

1.00 0.06 -0.29 -0.13 -0.47 -0.19 0.20 -0.10 -0.20 0.15 0.07 -0.01 -0.05

1.00 -0.18 -0.03 0.00 0.00 -0.07 0.02 -0.10 0.14 0.00 0.03 0.00

1.00 0.16 0.34 0.09 -0.09 0.07 0.23 -0.26 -0.06 -0.09 0.02

1.00 0.18 0.03 0.07 0.02 0.04 -0.09 -0.03 -0.04 0.07

1.00 0.18 -0.22 0.06 0.24 -0.22 -0.04 -0.01 -0.01

Table 4. Marginal effects after probit models to explain active mobility (M1), public mobility by bus (M2), and private mobility by car (M3) respectively. Source: own elaboration. Determinants Age Female gender Student role Distance Economic level Bike ownership Car ownership F: travel time F: weathercondition F: flexible schedule F: comfort F: economicfactor F: risk of theft F: environment

M1: ACTIVE MOBILITY dy/dx p-value -0.0044 0.100 ** -0.1223 0.003 *** -0,0783 0.219 -0.0262 0.000 *** -0.0093 0.796 +0.0991 0.018 *** -0.2910 0.000 *** -0.0350 0.397 +0.0893 0.049 *** -0.0619 0.139 * -0.0941 0.038 *** -0.0502 0.288 +0.0123 0.858 +0.2516 0.162 *

M2: PUBLIC MOBILITY dy/dx p-value -0.0077 0.037 *** +0.1096 0.008 *** +0.2074 0.001 *** +0.0189 0.007 *** -0.0104 0.782 -0.0312 0.481 -0.3310 0.000 *** +0.0467 0.343 -0.0025 0.961 +0.0389 0.473 -0.1177 0.028 *** +0.0587 0.350 -0.0719 0.236 -0.0426 0.757

M3: PRIVATE MOBILITY dy/dx p-value +0.0117 0.003 *** -0.0098 0.854 +0.0775 0.362 +0.0048 0.581 +0.1078 0.031 *** +0.0142 0.798 +0.5520 0.000 *** +0.0919 0.124 * -0.0757 0.224 -0.0014 0.983 -0.0120 0.878 -0.0892 0.194 * +0.1533 0.164 * -0.1974 0.124 *

Notes: estimations with robust standard errors; dy/dx calculated with x (continuous) and x = 0 (categorical). P-value: * means p < 0.20, **p < 0.10, ***p < 0.05 (dark blue shading)

5 Final Remarks The results help to understand how the analyzed university community moves and the determinants of mobility decisions. The outcomes show that private combustion vehicles (41–50%), public transport by bus (29–31%), and active mobility (20–26%) are the most used means of transport, with electric mobility being less than 2%. Regarding mobility determinants, we find that age and car ownership are relevant factors of the three types of mobility, being private mobility more used by older people

Understanding Urban Mobility Habits and Their Influencing Factors

121

with usually have a vehicle. Also, we identify factors related uniquely to each type of mobility: bike ownership and wheatear conditions—positive effect in active mobility (M1); student role -positive effect in public mobility (M2); and economic level -positive influence in private mobility (M3). Moreover, we find that active and public mobility share three determinants namely: female gender and distance -negative impact for active mobility and positive effect for public transport- and comfort -negative influence in both active and public mobility. These results allow us to risk a hierarchy of factors influencing mobility at three levels. First, age and car ownership determine the use of private mobility. Second, given the absence of previous factors, bike ownership and weather conditions determine the use of active mobility. Third, the student role -without own car or bike- determine use of public mobility. It is an original and value-added article for several reasons. First, it studies a university campus in a medium-sized city in an emerging country (few similar studies have been found for Latin America and medium-cities). Second, the richness of triangulating multiple primary and secondary data sources, interpreted from an interdisciplinary perspective, gives robustness to the results. Finally, the use of multivariate models to find mobility decisions determinants’ is a step ahead of other descriptive studies. As limitations, we identify the difficulty in generalizing the results and the differences in contexts and models with respect to other research to compare. However, despite being a case study, the research yielded results consistent with other findings from the literature, which reinforces the replicability of the study and its contributions. The main research implications are for public policymakers, to design incentives to promote habits that contribute to sustainable development. We can also contribute from the universities, for example by raising awareness of the environmental impacts generated by mobility, and providing infrastructure for active mobility, among others. Several lines of research have emerged from this work. In the short term, incorporate other variables into the analysis such as frequency of campus attendance and mobility schedules. Subsequently, generate mobility proposals that promote sustainable development and improve the problems detected in the diagnosis, involving the multidimensional aspects of the phenomenon and the citizens’ perception. And last but not least, we seek to implement some of the proposals in a pilot test phase. Acknowledgments. We would like to express our gratitude to the Inter-American Organization for Higher Education (IOHE) for providing the funding for this applied research project (code 029, item 2 in https://oui-iohe.org/es/ies-investigacion-colaborativa). Additionally, we extend our thanks to Universidad Nacional del Sur and CONICET for their support.

References 1. Maciejewska, M.W., Miralles-Guasch, C., Marquet Sardà, O.: Perfiles de población y uso de los transportes motorizados: evidencias desde el campus de la Universitat Autònoma de Barcelona (UAB). Documents d’anàlisi geográfica 66(3), 629–648. (2020). https://doi.org/ 10.5565/rev/dag.553 2. Buehler, R., Hamre, A.: An examination of recent trends in multimodal travel behavior among American motorists. Int. J. Sustain. Transp. 10(4), 354–364 (2016). https://doi.org/10.1080/ 15568318.2014.945672

122

G. Pesce et al.

3. Eastman, J.A.S. El uso de la bicicleta como promotor de la movilidad sostenible: acciones y efectos en la movilidad cotidiana, el mejoramiento de la calidad del aire y el transporte público de las ciudades. Revista Kavilando 12(1), 118–126 (2020). http://kavilando.org/rev ista/index.php/kavilando/article/view/381 4. Zapata-González, L., Quiceno-Hoyos, A., Tabares-Hidalgo, L. Campus universitario sustentable. Revista de Arquitectura 18(2), 107–119. (2016). https://doi.org/10.14718/RevArq. 2016.18.2.10 5. Lopera Quiroz, C., Lopera Calle, M.., Duque Quintero, D.A.: La universidad verde: percepciones de la comunidad universitaria en el proceso de transformación hacia la sostenibilidad. Revista Virtual Universidad Católica del Norte (57), 157–174. (2019). https://doi.org/10. 35575/rvucn.n57a11 6. González Calderón, C.A., Moreno Palacio, D.P., Gallón, S.V.: Análisis de la movilidad en campus universitarios: caso de estudio universidad de Antioquia. Revista Politécnica 7(12), 49–57 (2011). https://revistas.elpoli.edu.co/index.php/pol/article/view/179 7. Saladié, Ò., Jurado, J.: La movilidad en el campus Vila-seca de la URV: propuestas para una movilidad más sostenible. Investigaciones Geográficas (Esp) (64),163–182 (2015). https:// www.redalyc.org/articulo.oa?id=17643402010 8. Witte, A. de, Hollevoet, J., Dobruszkes, F., Hubert, M., Macharis, C.: Linking modal choice to motility: a comprehensive review. Transp. Res. Part A: Policy Pract. 49, 329–341 (2013). https://doi.org/10.1016/j.tra.2013.01.009 9. Tyrinopoulos, Y., Antoniou, C.: Factors affecting modal choice in urban mobility. Eur. Transp. Res. Rev. 5(1), 27–39 (2012). https://doi.org/10.1007/s12544-012-0088-3 10. Marquet, O., Miralles-Guasch, C.: Walking short distances. The socioeconomic drivers for the use of proximity in everyday mobility in Barcelona. Transp. Res. Part A: Policy Pract. 70, 210–222. (2014) https://doi.org/10.1016/j.tra.2014.10.007 11. Ahmad, S., de Oliveira, J.A.P.: Determinants of urban mobility in India: lessons for promoting sustainable and inclusive urban transportation in developing countries. Transp. Policy 50, 106–114 (2016). https://doi.org/10.1016/j.tranpol.2016.04.014 12. Zhou, J.: From better understandings to proactive actions: housing location and commuting mode choices among university students. Transp. Policy 33, 166–175 (2014). https://doi.org/ 10.1016/j.tranpol.2014.03.004 13. Delmelle, E., Delmelle, E.: Exploring spatio-temporal commuting patterns in a university environment. Transp. Policy 21, 1–9 (2012). https://doi.org/10.1016/j.tranpol.2011.12.007 14. De las Rivas Sanz, J.L., Escudero, F.I., Soto, J.L.L.: Campus universitario de Valladolid. Integración urbana y movilidad. Revista Bitácora Urbano Territorial 18(1), 139–156 (2011). https://revistas.unal.edu.co/index.php/bitacora/article/view/20640 15. Miralles-Guasch, C., Domene, E.: Sustainable transport challenges in a suburban university: the case of the Autonomous University of Barcelona. Transp. Policy 17(6), 454–463 (2010). https://doi.org/10.1016/j.tranpol.2010.04.012 16. Dávila, J.A.G., Canales, K.S.: Gobernanza y movilidad urbana hacia la sustentabilidad. Comunidad educativa en Monterrey, México. Revista Bitácora Urbano Territorial 30(3), 95–107 (2020). https://doi.org/10.15446/bitacora.v30n3.80196 17. Solórzano-Barreto, S.S., Villegas-Gorozabel, E.A., Delgado-Gutiérrez, D.A., MacíasSánchez, L.K.: Integración de una ciclovía en la movilidad interna de la Universidad Técnica de Manabí, Portoviejo. Revista Científica INGENIAR: Ingeniería, Tecnología e Investigación 5(9), 18–37 (2022). https://doi.org/10.46296/ig.v5i9edespjun.0066 18. Sancho-Almeida, I.S., Chiguano-Crespo, W.F.: La incidencia del campus disperso en la movilidad de los estudiantes de la Universidad Nacional de Chimborazo (Bachelor’s thesis, Riobamba: Universidad Nacional de Chimborazo) (2021). http://dspace.unach.edu.ec/ handle/51000/7253

Understanding Urban Mobility Habits and Their Influencing Factors

123

19. Castañeda, Á., Ocampo, G., Sánchez, K., Perén, J.: Movilidad urbana en el Campus Central de la Universidad de Panamá: Caso de la Facultad de Arquitectura y Diseño y la Facultad de Ciencias de la Educación. Revista de Iniciación Científica 4, 84–91 (2018). https://ridda2. utp.ac.pa/handle/123456789/9882 20. Córdoba Naranjo, I.A., García Amador R., Moll Martínez, R., Fundora Ayuso, G.: Caracterización y propuesta de medidas para mejorar la accesibilidad a la Universidad Tecnológica de La Habana. Revista de Arquitectura e Ingeniería 14(2), 1–8 (2020). https://www.redalyc. org/articulo.oa?id=193963490004 21. Soria-Lara, J.A., Marquet, O., Miralles-Guasch, C.: The influence of location, socioeconomics, and behaviour on travel-demand by car in metropolitan university campuses. Transp. Res. Part D: Transp. Environ. 53, 149–160 (2017). https://doi.org/10.1016/j.trd.2017.04.008 22. Bilbao-Ubillos, Fernández-Sainz: The influence of quality and price on the demand for urban transport: the case of university students. Transp. Res. Part A: Policy Pract. 38(8), 607–614 (2004) https://doi.org/10.1016/j.tra.2004.04.004 23. Hernández Sampieri, R., Fernandez Collado, C.: y Baptista Lucio. McGraw-Hill, P. Metodología de la investigación (2014) 24. Flick, U.: Qualidade na pesquisa qualitativa. (Traductor Costa, R.). Artmed (2009) 25. Instituto Nacional de Estadística y Censos. Censo Nacional de Población, Hogares y Viviendas 2022. Resultados provisionales (2023). https://www.censo.gob.ar/index.php/datos_provision ales/ 26. Cátedra de Agrometeorología, Universidad Nacional del Sur. Caracterización Climática de Bahía Blanca (2007). https://agrometeorologia.criba.edu.ar/Downloads/Climatic.pdf

A Conversational Agent for Smart Schooling A Case Study on K-12 Dropout Risk Assessment Renata Magalh˜ aes1 , Bruno Veloso1 , Francisco S. Marcondes1 , Henrique Lima2 , Dalila Dur˜ aes1(B) , and Paulo Novais1 1

ALGORITMI Centre/LASI, University of Minho, Braga, Portugal [email protected], [email protected], [email protected], [email protected], [email protected] 2 Codevision, S.A., Braga, Portugal [email protected]

Abstract. The goal of smart education is to utilize advanced technology in order to improve the teaching experience by establishing a stimulating and interactive atmosphere for learning. Conversational agents emerge as an aid for a smarter education. One of the possibilities to be explored is the building of tools that help predict and prevent student failure or dropout. This case study presents a research project that consists on the creation of a school platform for student interaction, in which a conversational agent, developed using Rasa, communicates with both the students and the class director and is able to assign a risk of academic failure, based on their answers to questionnaires scripted by a team of psychologists. XGBoost outperfomed AdaBoost, Decision Tree and Random Forest algorithms with an accuracy of 97%. Keywords: Rasa · Academic Failure Learning · Conversational Agent

1

· Smart Education · Machine

Introduction

Smart cities aim to enhance sustainability, increase efficiency and improve the overall quality of life for citizens. These cities integrate information and communication technologies (ICT) and Internet of Things (IoT) devices which collect and analyze data from numerous sources and that data is later used to manage assets and resources efficiently, make informed decisions and provide better services to citizens [1]. In this context, smart schooling emerges with the goal of using digital tools to improve the teaching and learning experience. It is a mean for an engaging and interactive learning environment. Conversational agents are a component of smart schooling as they can be a useful tool for building the student’s tailored profile and contributing to their needs by providing information and assistance. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 124–134, 2023. https://doi.org/10.1007/978-3-031-36957-5_11

A Conversational Agent for Smart Schooling . . .

125

Recently, the issue of using conversational agents in educational environments increased with the emergence of chatGPT [2]. Among matters related to plagiarism and academic honesty, which are surpassed by benefits to both students and professors [3,4], such as increasing student engagement and the facilitation of asynchronous communication [5], it may also help prevent school failure as proposed in this article. However, in spite of this common agreement about the usefulness of conversational agents for educational purposes, the majority of research has been carried out with higher education or secondary students. Information on the profitability of these agents in the context of compulsory education is lacking. Ethical Statement. Considering that the children are obviously underaged, the project and interaction plans were presented to their parents, and those who agreed were asked to sign an informed consent form. The scripts to be used by the conversational agent were written by a team of educational psychologist researchers who are working on the project alongside computer science researchers and a technology company. Further information about the team can be found on the project description subsection of this article. Students are aware that they are interacting with a chabot as they see an avatar and the agent presents itself in the first interaction. The class director and a psychologist are always present in the classroom at every testing moment. It is also important to note that the children are only able to interact with the chatbot in planned and controlled testing moments and never on any other occasions at the present time. Document Structure. The case study report is divided into three main sections. First, a project description is provided. Additionally, a component view is shown. Subsequently, the Education Intelligence and Conversational Agent modules are presented. This last section briefly refers to two examples of Rasa-based tutors in educational contexts. Following this, the outcomes are examined and a discussion is conducted, culminating in a conclusion.

2

The Case Study Report

A case study is a method to be used in this context, as, per definition, it provides thorough comprehension of a process, program, event, or activity [6] . These indepth insights can generate hypotheses for future research [7]. The object of this case study report is to address a K-12 dropout risk assessment using a dialoguebased structure. The approach used is descriptive. 2.1

Project Description

This project seeks to collaborate with the public government initiative of avoiding school failure and dropout for K-12 students. The target is to create a set of warnings that help the class director recognize the dropout risk and enable them to act in a timely manner.

126

R. Magalh˜ aes et al.

The team working on the project is multidisciplinary, consisting of: educational psychologist researchers, who are building the scripts, ensuring that the language used and conversation flow is appropriate for children of different ages and at different stages or their academic journey. These scripts will be followed by a team of computer science researchers in order to build the conversational agent. Meanwhile, the technology company, that works on developing software solutions in the Education and Training sectors, is in charge of the deployment of the agent. Together, they are leveraging the strengths of each discipline to create a digital platform that will help identify those who may be at risk of academic failure and dropout. 2.2

Component View

Overall, the proposed solution is composed of two modules, Education Intelligence and Digital Assistant, which will interact with three involved actors (class director, student and the school platform), as represented in Fig. 1.

Fig. 1. Components diagram representing the platform integration with the involved actors.

The Education Intelligence component is responsible for processing the information obtained from both the school platform and the student in order to produce information of interest to the class director but also data that help refine the behavior of the Digital Assistant module. The Digital Assistant module serves as an interaction support for the student. The student provides subjective information about how they feel about their school experience while the school platform provides objective information such as grades and absences.

A Conversational Agent for Smart Schooling . . .

127

The primary purpose of this platform is to support the class director with monitoring school failure or dropout risk, which is possible through the combination of all these components. 2.3

Education Intelligence

Machine Learning was utilized to forecast a student’s academic situation, which may have an impact on how to provide assistance to the student. To make this forecast, the Machine Learning model relied on the recent grades of the student in the subjects of Portuguese and Mathematics. Furthermore, in addition to these prerequisites, it also required the responses given by the students to the questionnaires from Rasa. Rasa is a widely used open source framework for building chat and voice-based AI assistants [8]. Prior to integrating the Machine Learning component with the Digital Assistant, it was necessary to investigate which algorithm was most suitable for classifying (and therefore predicting) a student’s academic situation. A team of psychologists created a survey for students to complete, which focused on their academic experiences. This survey was administered to students in different school years. Furthermore, in addition to the survey responses, the psychologists also had access to the students’ grades in Portuguese and Mathematics for both the first and second terms. Moreover, based on each student’s academic situation, the psychologists assigned a cluster to them. This cluster identifies the academic situation of a particular student, see Table 1. The dataset consisted of 845 entries and 123 columns. Among these entries, 422 were from primary school students and 423 were from middle students. Additionally, the dataset had a few null values and some missing data, so an initial pre-processing step was necessary. The distribution of students among the clusters can be seen in Fig. 2. It can be seen that the different clusters have a similar number of students. It is also noted that one fourth of the total number of students are classified as being at high risk of academic failure, which can be a concerning number at first glance. However, the school from which this data was collected is known for its ability to deal with struggling students due to the fact that it is a semi-private school located in a remote area, where students struggle to find distractions, so Table 1. Descriptions of each cluster. Cluster

Description

Cluster 0

Indicates that the student has a high risk of academic failure

Cluster 1

Suggests that the student is at a moderate risk of not succeeding academically

Cluster 2

Indicates that the student is at a moderate level of risk, although less serious than cluster 1

Cluster 3

Indicates that the student has a low level of risk and is therefore on a good path towards academic success

128

R. Magalh˜ aes et al.

it is possible that many of these students were transferred from other schools with this purpose, therefore explaining the high number of students at risk.

Fig. 2. Distribution of students among the clusters. See Table 1 for detailed explanation of each cluster.

Three studies were carried out, one for each dataset. The first dataset contained all the columns from the initial dataset. The second dataset included the 13 columns that psychologists had identified as the most relevant for classifying the student’s cluster. These columns included the student’s most recent grades in Portuguese and Mathematics, their satisfaction with their grades, their perception of whether their grades reflected their abilities, their level of engagement in school, their short and long-term selfregulation skills, their motivation (two columns), their perception of the value of education, and their support from family, teachers, and peers. The third dataset consisted of five columns. The columns are: the student’s most recent grades in Portuguese and Mathematics, their satisfaction with their grades, their perception of whether their grades reflected their abilities, and their level of engagement in school. Four algorithms were tested in addition to these three datasets. All of these algorithms had previously been studied as the most powerful in terms of academic situations [9]. The AdaBoost, DecisionTree, RandomForest, and XGBoost algorithms were applied. A Grid Search Cross Validation was applied to optimize the hyperparameters of each algorithm, and the datasets were divided into 80% training and 20% testing in all algorithms. Based on the results presented in Table 2, we can conclude that the highest accuracy is achieved when using the XGBoost and Random Forest algorithms. The XGBoost algorithm achieved an accuracy of 0.9704 in both dataset 2 and dataset 3, while the Random Forest algorithm

A Conversational Agent for Smart Schooling . . .

129

achieved its highest accuracy (0.9586) when using dataset 3. Generally speaking, dataset 3 performs better, and the best model for this dataset is XGBoost. Given these results, and in order to properly classify each student’s cluster, it was identified that it is only necessary to use the information present in dataset 3 and the XGBoost algorithm. Table 2. Results for each algorithm. Algorithm AdaBoost

2.4

Dataset 1 Dataset 2 Dataset 3 0.8698

0.8580

0.8817

Decision tree

0.8817

0.9231

0.9527

XGBoost

0.9645

0.9704

0.9704

Random forest

0.8935

0.9527

0.9586

Conversational Agent

Alongside Rasa there are other platforms that can be used for building chatbots, such as A.L.I.C.E. based chatbots (such as Pandora-bots). However, they differ in their features, approach and capabilities [10]. Rasa uses state-of-theart Transformer-based architecture for abducting complex relationships between words [8]. DIET (Dual Intent and Entity Transformer) is a neural network architecture that handles both intents and entity extraction. It is able to learn with the tokens and sentence characteristics and is considerably faster to train and parallels large-scale pre-trained language models in performance [11]. On the whole, Rasa is known for its Natural Language Processing versatility and learning capabilities, enhancing the “pattern matching” by extrapolating from instances. For that reason it is considered a good choice [12] considering this project’s constraints. Therefore, the conversational agent used in this research project was developed using the Rasa framework. For a reference, ArgueTutor is a Rasa-based tutor providing guidance on the writing process. It offers students adaptative and immediate feedback, theoretical knowledge and detailed guidance through their writing process. These features are integrated using rule-based trained chat intents following the architecture of Rasa NLU and Rasa Core. It has been tested by 55 students and researchers verified that by using ArgueTutor students wrote more convincing texts [13]. Another Rasa-based tutor embodied into a robot that helps children washing hands is worth of mentioning. This tutor was designed to to promote positive behavior change in children, specially in the task of hand hygiene. The robot gives feedback in real time to the children, to motivate them to maintain the quality of their hand-washing if they are doing it correctly, or to correct them in case they are not. And as for the conversational part, the robot can have

130

R. Magalh˜ aes et al.

conversations about hand hygiene and domain specific subjects, and for that it uses a combination of RASA, Google’s Automatic Speech Recognition and Text-To-Speech engines [14]. Rasa Framework The general structure of a Rasa project consists of a rasa NLU, rasa rules, rasa stories, actions and domain.yml.

Fig. 3. NLU file.

– NLU: The NLU (Natural Language Understanding) is responsible for understanding the user’s input and extracting relevant information such as user intent and entities. – Rules: The rules file is used to define conditional statements that help to control the flow of a conversation. – Stories: Rasa stories are pre-defined sequences of actions that represent a particular conversation flow. Stories allow developers to map out different conversation paths that the chatbot can take based on the user’s input. – Actions: Rasa actions are the specific tasks that the chatbot can perform in response to the user’s input. Actions can include sending a message, making an API request, or accessing a database. – Domain: The domain file defines the chatbot’s vocabulary, actions, and responses. It includes all the intents, entities, actions, and templates that the chatbot needs to understand and respond to user input. For this project there are a total of 502 actions, 63 intents, 10 rules and 4 stories. Some actions are simple responses such as “hello” and “goodbye”. The chatbot is also able to detect inappropriate words and redirect the conversation in case such words are detected, Moreover, it is also programmed to tell the students to ask the class director for help in case of confusion.

A Conversational Agent for Smart Schooling . . .

131

Since the target user of this conversational agent is K-12 students, it was decided that button-type answers as well as the use of emojis would be predominant. This means that in several interactions the user only has to select the button that matches their opinion or intention, for instance, in the questionnaire that is presented in the first activity, where the student has to select which answer applies most to the sentence shown, and the answers vary from “totally disagree” to “totally agree”. When the chatbot asks the students how their week has been, a set of emojis varying from happy to sad are presented. What is more, the chatbot has been developed in two languages: portuguese and english. Both versions share exactly the same characteristics and follows the same flow. The jokes presented by the chatbot were made sure to be equivalent and the language register is equal. 2.5

Realworld Experiments

A total of three activities have been fully developed and tested. The first activity starts with the conversational agent presenting itself and asking the students to introduce themselves and say what their eye and hair colour is, as well as asking them for a characteristic that they like about themselves or others like about them, in order to establish a rapport with the students. It then moves on to a questionnaire also written by the team of psychologists involved in the project, which included asking students for their opinion regarding their own study habits (e.g., “I always hand in my homework on time” and “After I finish my homework I revise to make sure it’s correct”), their view of education (e.g., “It’s important for me to learn as much as I can” and “It’s important for me to go to university”) and their family and friends support (e.g., “My parents are interesting in knowing about what goes on at school” and “My friends respect what I have to stay”). Then, the chatbot asks the students for their favorite free-time activities and gives information about how that activity can help them become a better student, for example, if a student says “doing sports”, the chatbot will explain how that activity might require team work, just like doing a group project at school. It was through this activity that the educational psychologists were able to assign a cluster to the students, based on their answers from the questionnaires, as explained in Sect. 2.3. In addition, this activity aims to show students that school can also be fun and they can take their daily routine and free time activities and apply their knowledge and strategies at school. In the second activity, the chatbot greets the students, asks them how their week has been and tells them a joke in order to make them feel comfortable and at ease. Afterwards, a context-situation is presented in which a colleague is struggling at school and the student is asked for strategies that this colleague can use in order to improve his grade. Finally, a few strategies are analyzed in detail so that the students can understand how they can improve their own performance at school. It is important to note that this activity performs based on the assigned cluster. If the student has previously been classified as a highrisk student, certain strategies were discussed, whereas different strategies were

132

R. Magalh˜ aes et al.

analyzed if the student has been classified as low-risk. The goal is to help the students in predicting consequences. The third activity consists of detailing these study strategies a bit further and showing possible outcomes of using or not using these strategies. What is more, a small exercise is also presented in which students have to try and organize a birthday party and, using these steps given by the students, the chatbot will explain how planning and doing things step by step can help them in the long term. This activity aims to use a real-life example to demonstrate to the students how they can apply these strategies at school in order to become better students and improve their grades.

3

Results and Discussion

Even though conversational agents can be a useful tool in educational contexts, information about its use in the field of education is scarce, especially at the ages of compulsory education. This also applies to the use of Rasa, which, in spite of its advantages, has not been fully explored in these contexts. School failure and dropout is a serious issue to which not many tools have been developed in regards to prediction and prevention. Conversational agents can be a useful tool in this sense. Therefore, this case study report observed and gathered insights on the impact of a conversational agent in regards to K-12 dropout risk assessment. Three tests were conducted in the first trimester of the present year and feedback was positive. Students adhesion was satisfactory and no major issues were reported. All tests were performed in a controlled environment in which the class director and a psychologist were present at all times. Parents of the children were previously informed and signed a consent form. Through the conversational agent, it is possible to obtain a student’s profile that refers to their risk perception. In order to do so, the student answers a questionnaire mediated by Rasa, and then these data go through a Machine Learning model that classifies the student. By predicting this information, it is possible to warn the class director who can then take action. Nonetheless, a few setbacks were encountered at the beginning of the study. These included different opinions regarding the type of answers that students could provide, as researchers deemed necessary to have more textual input in order to correctly perform sentiment analysis and personality detection but psychologists reckoned that children at the ages of primary and secondary education do not have an entirely developed personality. What is more, psychologists focused mainly on the safe conduct of the experience for the children, arguing and inferring appropriate conversation flows whereas computer science researchers’ focal point ranges from data collection to data analysis and techniques used, which were limited for the sake of finding common ground for the project. In order to ensure that the interactions with the conversational agent were engaging and enjoyable it was decided that emojis and button-type answers would be often available which hindered the textual input necessary for a proper analysis.

A Conversational Agent for Smart Schooling . . .

133

Overall, the use of a conversational agent in such a platform has been favourable and the multidisciplinarity profitable. Children find this tool entertaining and engaging, which enables retrieval of important data much needed in order to build a Decision Support System (DSS) that will help detect chances of school failure or dropout in advance. Additionally, the use of this type of tools can be advantageous considering that it may help the teacher to keep up with unnoticed or hidden situations. By signaling students at risk in a timely manner, it is possible to warn the class director and further action can be taken, resulting in an accurate prediction and prevention of school failure and dropout. Future work can include focusing on what type of action can be taken once risk is predicted.

4

Conclusion

Smart education aims to use digital tools to improve the teaching experience by creating an engaging and interactive learning environment. In order to achieve intelligent education, conversational agents are an important factor as they provide a more interactive environment and, through the construction of the student’s profile and other data, provide assistance to students when needed. However, the use of Rasa in educational contexts has not been fully explored, which was concluded by the difficulties in finding information on the topic. This case study concluded that it is possible to obtain an accurate student’s profile through the use of a conversational agent. It was found that XGBoost outperformed all other algorithms with an accuracy of 97%. Now, it is up to future work to develop a set of rules so that decisions can be made according to the student’s academic situation. These rules may include actions such as notifying the teacher or the parents, among other actions aimed at preventing academic failure. In the future, the plan is to develop a more comprehensive conversational agent that can provide broader responses and assess the sentiment and motivation behind them. Acknowledgments. This work is supported by: FCT—Funda¸ca ˜o para a Ciˆencia e Tecnologia within the RD Units Project Scope: UIDB/00319/2020 and the Northern Regional Operational Programme (NORTE 2020), under Portugal 2020 within the scope of the project “Hello: Plataforma inteligente para o combate ao insucesso escolar”, Ref. NORTE-01-0247-FEDER-047004.

References 1. Javed, A.R., Shahzad, F., ur Rehman, S., Zikria, Y.B., Razzak, I., Jalil, Z., Xu, G.: Future smart cities requirements, emerging technologies, applications, challenges, and future aspects. Cities 129, 103794 (2022) 2. Shaji George, A., Hovan George, A.S.: A review of chatgpt ai’s impact on several business sectors. Partn. Univ. Int. Innov. J. 1(1), 9–23 (2023)

134

R. Magalh˜ aes et al.

3. Resch, O., Yankova, A.: Open knowledge interface: a digital assistant to support students in writing academic assignments. In: Proceedings of the 1st ACM SIGSOFT International Workshop on Education through Advanced Software Engineering and Artificial Intelligence, pp. 13–16 (2019) 4. Mathew, A.N., Rohini, V., Paulose, J.: NLP-based personal learning assistant for school education. Int. J. Electr. Comput. Eng. 11(2088–8708), 4522–4530 (2021) 5. Cotton, D.R., Cotton, P.A., Shipway, J.R.: Chatting and cheating. Ensuring academic integrity in the era of ChatGPT (2023) 6. Yin, R.K.: Applications of Case Study Research. Sage (2011) 7. Newcomer, K.E., Hatry, H.P., Wholey, J.S.: Handbook of Practical Program Evaluation. Wiley Online Library (2015) 8. Rasa.: Introduction to rasa open source & rasa pro (2023). www.rasa.com/docs/ rasa/. Accessed 03 July 2023 9. Veloso, B., Barbosa, M.A., Faria, H., Marcondes, F.S., Dur˜ aes, D., Novais, P.: A systematic review on student failure prediction. In: International Conference in Methodologies and Intelligent Systems for Technology Enhanced Learning, pp. 43–52. Springer (2023) 10. AbuShawar, B., Atwell, E.: Alice chatbot: trials and outputs. Computacion y Sistemas 19(4), 625–632 (2015) 11. Bunk, T., Varshneya, D., Vlasov, V., Nichol, A.: Diet: lightweight language understanding for dialogue systems (2020). arXiv:2004.09936 12. Devi, C.P., et al.: Banking chatbot (b-bot. Turk. J. Comput. Math. Educ. (TURCOMAT) 12(10), 5795–5804 (2021) 13. Wambsganss, T., Kueng, T., Soellner, M., Leimeister, J.M.: Arguetutor: an adaptive dialog-based learning system for argumentation skills. In: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, pp. 1–13 (2021) 14. Prabha, P., Sasidharan, S., Pasupuleti, D., Das, A., Manikutty, G., Sharma, R.: A minimalist social robot platform for promoting positive behavior change among children. In: ACM SIGGRAPH 2022 Educator’s Forum, pp. 1–2 (2022)

Evaluation of Smart Mobility Indicators in Latin-American Cities Eladio E. Martinez Toro(B) , Elian Krut Yalil, and Vit Bubak Universidad Paraguayo Alemana, San Lorenzo, Paraguay [email protected]

Abstract. Major metropolitan cities in Latin America are facing transportation crises in terms of congestion, long travel times, pollution, accidents, and user safety. Similarly, to cities in developed countries, many cities in Latin America have implemented smart mobility programs to address these problems. In recent years, worldwide models have been developed to evaluate these programs, but most of these models use indicators and metrics that do not consider the mobility problems in Latin America. This research identified eight specific mobility challenges in Latin America and proposed new indicators with their metrics that allowed evaluating these challenges and the current state of smart mobility in the region. In addition, seven indicators with smart features were identified. The evaluation of the proposed indicators shows that the best positioned cities in Latin America in terms of mobility are Buenos Aires and Santiago de Chile. It also shows that the proposed model to evaluate mobility does not coincide with international rankings that evaluate mobility and transportation in Latin America, which makes it necessary to develop models that evaluate the reality of smart mobility in the region. This work will provide useful information for the development of a new mobility model with a more realistic and data-driven comparative assessment of mobility challenges and smart mobility programs in Latin American cities. Keywords: Smart mobility · Smart features indicators · Mobility challenges in Latin America

1 Introduction When we talk about the concept of mobility, we refer to the time and cost required to move an object, good or person from one place to another [1]. In recent times the concept of mobility has taken off due to the entry of new actors. One of these actors is Information and Communication Technologies or ICT, which have contributed to improve the ability, capacity and sustainability of the services offered by cities [2]. The implementation of these tools has given rise to the concept known as smart cities. A smart city is one that applies technology in different areas in a resource-efficient manner, either by saving energy, improving citizen services, or promoting the responsible and sustainable use of resources [3]. When evaluating a smart city, six fundamental pillars are considered. These are smart people, smart economy, smart governance, smart environment, smart living, and smart mobility [4], being this last pillar the one on which we will focus our research. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 135–146, 2023. https://doi.org/10.1007/978-3-031-36957-5_12

136

E. E. Martinez Toro et al.

When we analyze the growth and development of different geographic regions of the world, we find that Latin America is the region with the highest population increase, characterized by high levels of economic and social inequality and rapid growth of urbanization, which translates into growing mobility problems [5]. Although some smart mobility and public transport integration projects have started to be developed in Latin America to address these problems, smart mobility is one of the pending tasks for the region [6]. Evaluating the progress in smart mobility in Latin America is complicated since most of the existing smart mobility models were developed with the existing problems in developed cities in mind. Consequently, there are currently no mobility models containing indicators that measure the state of smart mobility in Latin American cities. Therefore, the objective of this research is to identify and evaluate the indicators that best describe the reality of smart mobility in Latin America and to determine which metrics best suited to evaluate these indicators.

2 Evaluation of Smart Mobility Models In the research carried out by Martinez-Toro et al. [7], eleven smart mobility models were evaluated which propose different indicators closely related to smart features. Of all the models evaluated, only four present a ranking that allows categorizing cities in terms of smart mobility. The model proposed by Giffinger et al. [4] presents a ranking of smart cities exclusively for medium-sized European cities. The models proposed by Battarra et al. [8] and Aletà et al. [9] evaluate specific cities in Italy and Spain respectively. Battarra’s model only focuses on mobility indicators while Aleta’s model evaluates mobility, environment, and smart city indicators. The only model that presents a global ranking that includes smart cities in Latin America is the model proposed by Berrone and Ricart [10] in the report IESE Cities in Motion Index 2022. This model evaluates the six dimensions that define a smart city. Regarding the indicators with smart features included in these proposed models, we can highlight that the most recurrent when evaluating smart mobility are clean energy transport, CO2 emissions control, use of smart cards, access to information in real time, integration of payment in different modes of transport, and the use and integration of ICT’s.

3 Mobility Challenges in Latin America In the work done by Martinez-Toro et al. [7] also evaluated the problems that affect mobility in Latin American cities. These problems were classified into three main categories: physical, infrastructure, and safety and environmental problems. For the purposes of this research, the problems, indicators, and some of the metrics proposed by Martinez-Toro et al. [7] were considered. We can mention that in the category of physical problems, the problems of traffic congestion and the limited availability of ICT’s were considered as part of this investigation. The indicators proposed to evaluate these problems were the traffic index and the number of vehicles per capita. Regarding the problem of ICT availability, the coverage of the Internet network in cities and the availability of digital platforms for transport services with real-time information were proposed as indicators. Regarding the infrastructure category, the problems evaluated were the integration of

Evaluation of Smart Mobility Indicators in Latin-American Cities

137

transportation and payment systems, as well as the availability of financial resources. The indicators proposed in this research for these problems are the availability of geographically integrated modes of transport, modes of transport with systems with integrated digital payment and the availability by users of a bank account that allows them to access mobility applications online. Lastly, for the safety and environment category, the identified problems were air pollution generated by transportation services using fossil fuels, the traffic accidents, and user safety. For this research, CO2 emissions, transport-related victims and public transport services that offer real-time information, respectively, are proposed as indicators. As part of the safety and environmental problems, due to its great relevance, it was decided to add an additional problem previously studied by MartinezToro et al. [6] which consists of the need for cities to implement mobility programs based on sustainable initiatives. This last problem allowed us to complete Table 1, which shows the relation between the problems, indicators, and metrics that most affect mobility in Latin America used in this research. Table 1. Proposed smart mobility indicators for Latin America. Mobility Problem

Indicator

Metrics

Vehicular Congestion

Traffic index [10] Number of vehicles per capita [12]

Composite index of time in minutes consumed in traffic due overall inefficiencies in the traffic system [11] Number of private vehicles per inhabitant [10]

ICT infrastructure

City-wide internet coverage [13] Digital transport service platforms with real time information [6]

Mobile-cellular subscriptions per 100 inhabitants [13] Active mobile-broadband subscriptions per 100 inhabitants [13] Percentage of transport companies with digital transport service platforms that offer real time travel information [6]

Transport system and payment integration

Geographically integrated transport modes [6] Integrated digital payment systems among transport modes [6]

Percentage of transport service companies with integrated transport modes [6] Percentage of transport service companies with integrated digital payment systems among transport modes [6]

Financial resources

Bank accounts [6]

Account ownership at a financial institution or with a mobile-money-service provider (% of population ages 15 +) [14] (continued)

138

E. E. Martinez Toro et al. Table 1. (continued)

Mobility Problem

Indicator

Metrics

Environmental pollution

Emissions of CO2 [10]

Estimation of CO2 consumption due to traffic time measured in grams for the return trip [12]

Traffic accidents

Victims related to transportation [15]

Mortality caused by road traffic injury (per 100,000 population) [15]

User safety

Public transport services that offer real-time information [12]

Percentage of public transport companies that offer on-line and real-time travel information [12]

Sustainability

Number of initiatives that enhance longer term sustainability of the public transport system [6]

Implemented sustainable mobility programs [6] Number of carpooling, car sharing, ride sharing, bike sharing programs [6]

Source: Own elaboration based on Martinez-Toro et al. [7]

For the purposes of this research, we will focus on the smart features that directly affect smart mobility. In the physical aspect we will analyze in detail city-wide internet coverage [13] and digital transport service platforms with real time information [6]. For the infrastructure category we will evaluate geographically integrated transport modes [6] and integrated digital payment systems among transport modes [6]. Finally, for the environmental and safety category we will consider emissions of CO2 [10], public transport services that offer real-time information [12] and number of initiatives that enhance longer term sustainability of the public transport system [6].

4 Methodology To develop this work, based on the article presented by Martinez-Toro et al. [7] which evaluated eleven smart mobility models and proposed the most relevant mobility problems in Latin America, the most appropriate indicators and metrics were selected to evaluate mobility in this region. It is important to highlight that part of this work was to identify indicators with smart features, which were detailed in the previous section. Once the indicators to be evaluated had been determined, the most relevant Latin American cities to be evaluated as part of this work were identified. To define the cities, we started with the cities previously evaluated in the work presented by Martinez-Toro et al. [6]. As part of this research, eight cities were evaluated in terms of the integration of their public transport systems. These were Mexico City, Bogota, Curitiba, São Paulo, Rio de Janeiro, Buenos Aires, Santiago de Chile, and Lima. All these cities were included in this paper. To expand the research, it was necessary to include additional cities, so the IESE Cities in Motion Index 2022 [10] report was selected, which, as mentioned above, is the only ranking that evaluates Latin American cities. This ranking considers

Evaluation of Smart Mobility Indicators in Latin-American Cities

139

a synthetic indicator that makes it possible to evaluate the effectiveness of cities in their smart practices. The indicator evaluates a series of dimensions and assigns a weight to each one of them. Based on the evaluation of 112 indicators in 183 cities, the index classifies cities as high, medium, or low. It is important to note that when reviewing the ranking, all Latin American cities are classified as medium or low. The best ranked city in Latin America is Santiago de Chile followed by Buenos Aires and completing the top five are Mexico City, Panama City and Montevideo [10]. To expand the study, additional cities were selected that were present in the IESE Cities in Motion Index ranking and were the top-ranked cities in their country. In the case of Argentina, we included the city of Rosario, since this city has made a very strong commitment in the region to make the city an example in terms of mobility and sustainability, following the path towards a smart city as set out in its Strategic Plan 2030 [16]. Once the problems, indicators with their metrics and cities were identified, we proceeded to evaluate the indicators in each of the cities. Secondary sources such as published articles, institutional reports, governmental websites, and specialized websites were used for this purpose. This allowed us to have a vision of the reality of smart mobility in Latin America. After concluding the evaluation of indicators, the cities were ranked in terms of their performance in each metric. Cities were ranked in ascending or descending order, depending on the criterion evaluated. For example, in the case of Digital transport service platforms in Fig. 2, which evaluates transport companies with digital transport service platforms that offer real time travel information, the city with more available platforms is better positioned than the one with fewer platforms. On the other hand, in the case of Fig. 5 Emissions of CO2 (in grams), which consists of the estimation of CO2 consumption due to traffic time measured in grams for a trip [11], the city with the lowest amount of CO2 grams is better positioned than the one with the highest amount of CO2 grams. Finally, once the evaluation of the indicator metrics was completed, a value was assigned to each city with respect to its position in the ranking and the values corresponding to each city in each metric were added up. This made it possible to establish an unweighted ranking that prioritizes the state of smart mobility in each city based on the proposed indicators.

5 Results and Analysis As mentioned in the previous section, the IESE Cities in Motion Index 2022 ranking evaluates different dimensions. One of these dimensions is mobility and transportation [10]. When evaluating the results for Latin America cities in this dimension, the result attracted much attention, so it was necessary to evaluate the indicators the IESE Cities in Motion Index 2022 used to evaluate this dimension. When analyzing the indicators, we can clearly see that they do not represent the reality of Latin America. For example, the indicator considers whether the city has programs related to the rental of bicycles, mopeds, and scooters. In addition, the indicator evaluates the number of bicycles per household and whether there are bicycle sharing programs in the city. Another aspect that is evaluated is the length of the subway network, the number of subway stations, whether the city has a high-speed train and the number of incoming flights (air routes) in a city. After reviewing in detail, we found that the indicator measures aspects that are

140

E. E. Martinez Toro et al.

not necessarily the reality of the region. This reaffirms the need to conduct this research and determine the current state of smart mobility and thus be able to determine the steps to follow to achieve its full development in the region. The following is a detailed evaluation of the smart features indicators separated into the three proposed categories. First, the indicators related to the physical category were analyzed. Below are the results of active mobile broadband subscriptions per 100 inhabitants and digital platforms for transportation systems metrics.

Fig. 1. Active mobile-broadband subscriptions per 100 inhabitants.

Fig. 2. Digital transport service platforms.

Figure 1 we can evaluate the active mobile-broadband subscriptions per 100 inhabitants, finding that Santiago de Chile leads the ranking with 111 per 100 inhabitants while Quito is at the bottom of the ranking with 57 per 100 inhabitants. Figure 2 shows the ranking of digital transport service platforms, where Sao Paulo leads the ranking with six platforms while at the bottom of the ranking are Bogota, Curitiba, Santiago de Chile, and Montevideo with only one platform. The second category evaluated was the one related to infrastructure problems. The following are the results of the evaluation of the metrics of geographically integrated transportation modes and Integrated digital payment systems among transport modes [6]. Regarding the indicators of the infrastructure aspect, Fig. 3 shows that Mexico City leads the ranking of geographically integrated transportation modes with nine integrated modes while at the bottom of the ranking are Montevideo and Asuncion, which have

Evaluation of Smart Mobility Indicators in Latin-American Cities

141

Fig. 3. Geographically integrated transportation modes.

8 7 6 5 4 3 2 1 0

Fig. 4. Integrated digital payment systems among transport modes

only one integrated transportation mode. Figure 4 shows that Mexico City is the city with the most integrated digital payment systems between modes of transportation with seven modes. In this case we have four cities that only have one mode of transport integrated to the digital payment system. Finally, indicators related to safety and environmental category were analyzed. The following are the results of the evaluation of the grams of CO2 emissions, public transport services offering real-time information, sustainable mobility program per city and the total of shared mobility platforms per city metrics. Regarding the safety and environment aspect, we find in Fig. 5 that Rosario leads the ranking with 2009.57 g of CO2 while the cities with the highest values are Asuncion and Mexico City with 8518.14 and 9452.31 g of CO2 respectively. Figure 6 shows the transport services that offer real time information, finding that Lima leads the ranking with five transport services while Santiago de Chile, Rosario and Montevideo only have one service with real time information. Figure 7 shows the sustainable mobility programs by city and shows that the ranking is led by Buenos Aires and Mexico City, since both cities have four initiatives. Rosario, Quito, Panama City and La Paz only have one initiative and in the cases of the last two, they only have long-term plans. Finally, Fig. 8 shows the total of shared mobility platforms per city where Mexico City is the

142

E. E. Martinez Toro et al.

Fig. 5. Emissions of CO2 (in grams)

Fig. 6. Transport services that offer real time information.

Fig. 7. Sustainable mobility programs per city.

city with the most carpooling, car sharing, ride sharing and bike sharing programs with 15 programs, while at the bottom of the ranking we have Panama City with only one shared mobility program. Table 2 shows the result of the unweighted ranking, which considers the position of each of the cities with respect to each of the metrics evaluated.

Evaluation of Smart Mobility Indicators in Latin-American Cities

143

Fig. 8. Total of shared mobility platforms per city.

Table 2. Unweighted ranking for the cities evaluated. City

Unweighted Ranking

Buenos Aires

64

Santiago de Chile

67

Lima

75

Rio de Janeiro

76

São Paulo

78

Mexico City

80

Montevideo

83

Curitiba

85

Rosario

92

Bogota

96

Panama City

104

La Paz

114

Quito

120

Asuncion

132

Source: Own Elaboration

As we can see in Table 2, the best positioned cities in the proposed ranking are Buenos Aires and Santiago de Chile, coinciding with the IESE Cities in Motion Index ranking. In the cases of both cities, although they only lead in one or two of the metrics evaluated, in most of them they remain in the top half of the ranking, which allows them to obtain a better positioning. In the case of Santiago de Chile, the most marked difference is in the metric that evaluates the digital platforms of transport services, since it is in the last position because the city has an integrated transport system, so that all information is on a single platform. At the bottom of the ranking, Asuncion coincides as the worst positioned city in both rankings. It is worth noting that Panama City, in comparison with the IESE Cities in Motion Index and the ranking of the mobility and

144

E. E. Martinez Toro et al.

transport dimension, is not at the top of the unweighted ranking. The main reason is because the proposed model does not evaluate the metro lines and stations or flights arriving and departing the city. Panama City currently has an efficient metro service and is also the Latin American hub for COPA Airlines. Due to this, we understand that Panama City is not well positioned in the proposed ranking. For this reason, it would be advisable to consider in the future including the metric of metro stations and lines, as eight of the cities evaluated in this study already have metro service. Another major difference compared to the IESE Cities in Motion Index are Lima and Rio de Janeiro, which improved their positioning because they are the top in one or two rankings while in the other their ranking was average. This makes both cities better positioned in the proposed model. As mentioned in Sect. 2, the only smart mobility model with a ranking that allows comparison of the results obtained in this research is the one proposed by Berrone and Ricart [10]. Therefore, it was not possible to make a comparison with the rankings proposed by other models. When comparing the results obtained in this work with the ranking of the mobility and transportation dimension of the IESE Cities in Motion Index, we see that the results obtained do not coincide. One of the main reasons is that the proposed ranking has not yet assigned a weighting to the indicators of the model. In addition, the proposed mobility model combines indicators that correspond to different dimensions of the IESE Cities in Motion Index, such as environment, mobility and transportation, urban planning, and technology. Regardless of the differences found in the ranking of the mobility and transportation dimension, all cities, including Asuncion, showed better performance when evaluating indicators that are present in all cities. Finally, we can mention that if we consider the state of the indicators with smart features present in Latin America, these are a pending task for most of the cities studied. Although we found quite advanced cities such as Mexico City and Buenos Aires, most of them are in the initial phase or in the planning process of related projects focused on the development of smart features that support smart practices.

6 Conclusions and Future Work This study aims to develop a new smart mobility model that best fits the reality of the Latin American region. To develop this research, first it was necessary to determine and evaluate the proposed smart mobility indicators in fourteen representative cities in Latin America using secondary sources. The cities considered as part of this study were selected considering previous works developed by the author [7] and the IESE Cities in Motion index ranking proposed by Berrone and Ricart [10], which contains a series of Latin American cities that were evaluated based on a global model of smart practices. It was also possible to determine which indicators with smart features are present in the region. These are city-wide internet coverage, digital transport service platforms with real time information, geographically integrated transport modes, integrated digital payment systems among transport modes, emissions of CO2 , public transport services that offer real-time information and the number of initiatives that enhance longer term sustainability of the public transport system. The proposed methodology allowed us to determine that the best positioned cities in Latin America in terms of smart mobility are Buenos Aires and Santiago de Chile. It

Evaluation of Smart Mobility Indicators in Latin-American Cities

145

was also demonstrated that the proposed model to evaluate mobility does not coincide with the worldwide ranking that evaluate smart mobility in Latin America. This makes it essential to develop a model that evaluates the existing reality in terms of smart mobility in the region. Finally, we can say that this work allows us to have a clearer vision of the reality of smart mobility and transportation systems in Latin America. To complete this research, it will be necessary to develop an index that weights the values of the proposed smart mobility indicators. To determine the relevance of the indicators, we will conduct a survey that will be disseminated to end users and to all stakeholders of public transport systems. The objective of this survey, in addition to measuring the relevance of the indicators, is also to know the opinion of stakeholders in public transport issues, such as transport companies, government representatives and researchers specialized in the area. It is proposed to use the experience and opinion of these people to determine first if the indicators are relevant; second if they really measure the challenges of smart mobility in Latin America; and third if they can be considered to develop a model of smart mobility unique to Latin America. With this information it will be possible to assign individual weights to each indicator, giving more weight to the most relevant indicators and less weight to the least relevant ones, and to develop a smart mobility index that specifically assesses existing problems in Latin America. This model will serve as a starting point for stakeholders to develop short-, medium- and long-term plans with the intention of promoting smart mobility in the region. Likewise, this work will help identify smart feature services and how they can improve mobility in the region’s cities.

References 1. National Research Council: Key Transportation Indicators: Summary of a Workshop. National Academies Press, Washington D.C. (2002) 2. Smart Cities Council: The definition of a smart city. https://www.smartcitiescouncil.com/art icle/2022-defining-smart-cities-again. Last Accessed 23 Feb 2023 3. Cohen, B., Obediente, E.: Study “Ranking of smart cities in Chile”. Estudio “Ranking de ciudades inteligentes en Chile” (2014) 4. Giffinger, R., et al.: Smart Cities: Ranking of European Medium-Sized Cities. Centre of Regional Science (SRF), Vienna (2007) 5. Hidalgo, D., Huizenga, C.: Implementation of sustainable urban transport in Latin America. Res. Transp. Econ. 40(1), 66–77 (2013) 6. Martinez-Toro, E.E., Van der Krogt, A., Sanchez-Flores, R.: Mobility and integration of public transport systems in Latin America. In: MLMI 2019: Proceedings of the 2019 2nd International Conference on Machine Learning and Machine Intelligence (2019). https://doi. org/10.1145/3366750.3366760 7. Martinez-Toro, E.E., Van der Krogt, A., Samaniego-Orue, C.G.: Towards a new model to evaluate smart mobility in Latin America. In: SMART 2021, The Tenth International Conference on Smart Cities, Systems, Devices and Technologies. ISBN: 978-1-61208-863-1 https:// www.thinkmind.org/index.php?view=article&articleid=smart_2021_1_40_40016 8. Battarra, R., et al.: Smart mobility in Italian metropolitan cities: a comparative analysis through indicators and actions. Sustain. Cities Soc. 41, 556–567 (2018) 9. Aletà, N.B., et al.: Smart mobility and smart environment in the Spanish cities. Transport. Res. Procedia 24 (2017)

146

E. E. Martinez Toro et al.

10. Berrone, P., Ricart, J.E.: IESE cities in motion index 2022. IESE, Business School University of Navarra, Spain (2022). https://media.iese.edu/research/pdfs/ST-0633-E.pdf. Last accessed 20 Feb 2023 11. Numbeo. About Traffic Indices at This Website. https://www.numbeo.com/traffic/indices_e xplained.jsp. Last Accessed 21 Feb 2023 12. Lupiañez, F., Fauli, C.: Smart cities: social evaluation of Smart Cities projects, Center for Telecommunications Studies of Latin America. Ciudades Inteligentes: Evaluación social de proyectos de Smart Cities, Centro de Estudios de telecomunicaciones de América Latina (2017) 13. International Telecommunication Union (ITU): Individuals using the Internet (% of population), World Telecommunication/ICT Indicators Database. https://data.worldbank.org/indica tor/IT.NET.USER.ZS. Last Accessed 23 Feb 2023 14. The World Bank: Data Bank, Account ownership at a financial institution or with a mobilemoney-service provider (% of population ages 15+). https://data.worldbank.org/indicator/FX. OWN.TOTL.ZS. Last Accessed 23 Feb 2023 15. The World Bank: Data Bank, Mortality caused by road traffic injuries (per 100,000 people). https://data.worldbank.org/indicator/SH.STA.TRAF.P5. Last Accessed 23 Feb 2023 16. Municipalidad de Rosario, Plan Estratégico Rosario 2030. https://www.rosario.gob.ar/Archiv osWeb/libro_rosario_2030.pdf. Last Accessed 22 Feb 2023

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation Daniel Montes Agudelo1 , Jheimer Julián Sepúlveda López1,2 and Luz Arabany Ramírez Castañeda1(B)

,

1 Universidad Nacional de Colombia, Manizales, Colombia

[email protected] 2 Universidad Nacional Abierta y a Distancia (UNAD), Bogotá, Colombia

Abstract. The terms ‘adoption’ and ‘appropriation’, when referring to information and communication technologies, are used indiscriminately. The differences between the two are not highlighted, nor are the objectives thereof established. Similarly, there are various model proposals which may explain the ICT adoption and appropriation processes in a given community. The diverse individual and group characteristics which enable said processes, as well as their implementation for the achievement of objectives and the encouragement of labor, educational, and political activity participation, were identified by way of a literature review. When addressing ICT adoption and appropriation quantitively, the characteristics therein seem independent. However, bibliographical analysis reveals their interrelation, together with those compositional variables which are heterogeneous and interdefinable. Based on the aforementioned postulate, this document reviews assorted ICT adoption and appropriation models, and a consensus is reached for both individual and group variables, as well as a relationship model which uses Interpretative Structural Modeling (ISM). Keywords: TIC adoption · Digital inclusion · TIC appropriation · ISM methodology

1 Introduction A review of the academic literature regarding information and communications technology adoption and appropriation processes permits the identification of multiple proposals for the characteristics that determine these processes. This process revealed that said terms are not defined and differentiated in the reviewed material. ICT adoption and appropriation are processes which are described by different means, variables, and characteristics. The identification of these descriptions and the construction of a consensus of their meanings and relationship permits digital inclusion to be addressed with a complex, multi-dimensional approach. The description of the digital inclusion phenomenon, through the simultaneous integration of social and technological aspects, permits a focus © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 147–157, 2023. https://doi.org/10.1007/978-3-031-36957-5_13

148

D. M. Agudelo et al.

on a sociocybernetics challenge: “to generate an interdisciplinary focus, so as to interact with the government in relation to complex social obstacles” [1]. Thus, the objective established for the present study is to construct a relationship model, with both individual and group characteristics, for ICT adoption and integration, which utilizes interpretative structural modeling. As mentioned previously, the review of the available literature on technology adoption and appropriation, demonstrated that authors use said terms indiscriminately, without providing definitions or indicating the existence of any difference between the two. Likewise, in the bibliographic material, these terms are employed without any explanation of the significance that they may have, in relation to ICT. The literature reports that these terms are both related and complementary. Schwarz and Chin [2] affirm that guaranteeing end-user acceptance of Information Technologies (IT) is an important challenge and for the administration thereof. For [3] technology acceptance forms part of a ladder, in which, basic access to technology is the bottom rung. The next rung up consists of the adoption and development of the basic requirements for technology use. The last rung is appropriation, which indicates strategic use, where an individual or organization uses ICT for their own purposes. Surman and Reilly [4] propose that adoption and appropriation are the steps of a stairway. The initial step is basic access: a computer connected to the internet, a cell phone with text messages, or a cybercafé. The second step is adoption and development of the basic abilities needed to use technology as intended. The final step is appropriation, or strategic use, in which an individual or organization converts the technology for their own ends and makes it their own. Appropriation includes putting local content on the internet in local languages or designing applications to meet an organization’s specific needs. It may be concluded from the literature review that technology appropriation refers to user interaction with technology, the changes in technology use which result therefrom, and the social framework in which it is used. Beyond that, technology appropriation is a cultural process, in which adoption and modification is intended to conform to the culture and norms of a particular group. Thus, appropriation refers to TIC design, personalization, and appropriation for purposes other than those originally foreseen. Here, the difference between use and appropriation occurs when technology is adapted and reflects cultural objectives. In practice, however, the process of appropriation follows a cyclical process involving interaction and exploration [5]. Camacho Jiménez [6] posits that an organization, country, or person has appropriated the internet when it has been incorporated fluidly into their daily activities, when they are able to discern the convenience of technological tools to solve daily problems, how they may be effectively combined with other instruments, and when it is possible to naturally establish national, organizational, or personal procedures, policies, and strategies to capitalize on the internet. This document is structured as follows: firstly, a review is performed of different ICT models, and proposals for ICT adoption and appropriation’s personal and group characteristics are identified. Secondly, the process implemented for establishment of characteristic consensus is detailed. Thirdly, the conclusions of the present investigation are presented.

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation

149

2 Models for ICT Adoption and Appropriation The authors reviewed the various models proposed, which describe ICT adoption and appropriation, based on sets of variables and relationships with those that explain them. Differences were identified, based on context, community, and the experts’ previous experiences. The characteristics employed for diverse purposes depended, among other considerations, on the type of situations analyzed. For some authors, group characteristics were more relevant, including, for example, the differences between countries or other geographic areas. Other authors proposed that personal characteristics, such as age, gender, income, and economics are factors that should be considered in the explanation of ICT adoption and appropriation. Table 1 lists a compendium of authors and their respective proposals for individual and group characteristics for ICT adoption and appropriation.

3 Consensus of Characteristics for ICT Adoption and Appropriation 3.1 Interpretative Structural Modeling (ISM) In order to achieve consensus and identify the relationship between the selected characteristics, it was decided to apply the Interpretative Structural Modeling (ISM) methodology, as it reveals the contextual relationship between a set of variables that conform a given issue or problem. In this case, the problem was the definition of the individual and group variable relationship between ICT adoption and appropriation, as identified in the literature review. The ISM methodology systematically applies certain elementary notions of graph theory, and exploits the theoretical, conceptual, and computational concept of graphs, so as to efficiently construct a directed graph with a complex pattern which corresponds to a contextual relationship between a set of elements. The systematic consideration indicates that, “for whichever complex problem being considered, a series of factors may be related with an issue or problem. However, both direct and indirect relationships among the factors describe the situation much more accurately than any individual factor taken in isolation” [7]. The different steps of the methodology are illustrated in Fig. 1. 3.2 Model Review and Consensus Building With the purpose of building a relationship model between individual and group variables, the variables proposed by each of the models, which were found in the reviewed articles, were identified. Table 1 summarizes the result of this process.

150

D. M. Agudelo et al.

Fig. 1. Flow chart for ISM model preparation. Source: [8] Table 1. Number of variables with and without definition. Article (model)

Authors

Number of variables

1

Sun y Zhang [9]

2

Noce y McKeown [10]

7

x

3

Maldifassi y Canessa [11]

3

x

4

Billon, Marco y Lera-Lopez [12]

5 6

10

With definition

Without definition

x

19

x

Schleife [13]

9

x

Weber y Kauffman [14]

6

x

7

Almuwil, Weerakkody y El-Haddadeh [15]

12

x

8

Gonzaléz, Sánchez y Galvis [16]

30

x

9

Kubiatko [17]

4

x

10

Taipale [18]

8

x (continued)

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation

151

Table 1. (continued) Article (model)

Authors

11

Pick, Sarkar y Johnson [19]

7

12

Hinostroza, Matamala, Labbe, Claro y Cabello [20]

6

x

13

Papaioannou y Charalambous [21]

8

x

14

Koç, Turan y Okursoy [22]

6

15

Thill, Rosenzweig y Wallis [23]

5

16

Nilashi, Ahmadi, Ahani, Ravangard y bin Ibrahim [24]

Total

Number of variables

With definition

Without definition

x

x x

17

x

157

86

71

The next step, once the variables were identified, was to classify individual and group variables. For this, the name of the variable, definition, and the context in which the author uses it were considered. Of the 157 variables identified, 91 variables were classified as individual, and 31 as group variables. The remaining 35 were not classified, and were discarded because, for the purposes of this study, they were not relevant for two reasons: • They are proposed and used to model specific cases, and not for the general applications which correspond to the context of this investigation. • The authors did not provide a definition of the variable, and its name is not clear enough to identify, if it belongs to the group of individual or group variables. Once the classification process for the variables was carried out, it was clear that some of these had the same name, or in some cases, the meaning or context in which they were used was the same. Based on this review, it was decided that similar variables could be unified. When completing the unification, for each one, the name that best represented the variable and its description was employed. At the end of this process, a total of 20 individual variables were obtained (Table 2) and 18 Group characteristics (Table 3).

152

D. M. Agudelo et al. Table 2. Individual characteristics for ICT adoption and appropriation.

Individuals

Description

Social class

Socioeconomic status

Education level

Highest level of education achieved

Age

Number of years

Residential area

Location of residence (urban/rural area)

Gender

Sexual identity, understood from the sociocultural point of view (female, male, for example)

Language

Native language (Spanish, French, for example)

Minors in the home

The existence of people under 18 years of age at home

Income

Quantity of money earned monthly

Marital status

Condition related to birth, nationality, parentage, or marriage (married, single, for example)

Occupation

Work and employment

Intellectual abilities

Intelligence level (low, medium, high)

Experience with ICT

Number of years that ICT products have been used

Race

Groups whose characteristics are perpetuated by inheritance (for example, white, indigenous, black)

Ethnicity

Community defined by racial, linguistic, cultural affinities, etc

Place of ICT use

Principal space in which ICT products are used (home, office, educational institution, for example)

Trust in ICT

Security and comfort in using ICT

Motivation

Interest in learning and openness to having new experiences

Disability

Physical or mental disability

Significant use of ICT products

Guidance on the use of ICT (job or training)

Community ICT trust

Degree to which the community feels safe and backs the use of IT

3.3 Expert Validation—Accessibility Matrix Table 4 lists the experts who participated in variable validation and in the process of establishing the relationships between them. The empty accessibility matrix was sent to each of the experts, and each one was to write ‘1’ if the row variable determined the column variable, or a ‘0’ if that relationship did not exist. In addition to the opinion of the experts, an accessibility matrix was built, which considered the 16 models that are reviewed and detailed above.

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation

153

Table 3. Group characteristics for ICT adoption and appropriation. Groups

Description

Institutional support

If there is public and/or private investment for ICT training

Culture

Common aspects that define a community

Legal norms

Laws or regulations related to the use of ICT products

Religion

Set of beliefs or dogmas about divinity, feelings of veneration shared by a community

Size of city

Number of inhabitants

Population density

Number of people per square kilometer

Urban population

Percentage of total location located in an urban area

School expectations

Maximum expectation of school years in the community

Perception of usability

The degree to which ICT products are considered applicable to different activities

Ease of use perception

The degree to which ICT products are seen to require minimal effort

Degree of imitation

Degree to which the individuals of a community imitate others

Size of home

Number of people at home

Willingness

Measure by which the adoption decision is perceived to be optional

Acceptance of innovation

Openness to new ideas

Age of population

Average population age

Compatibility

Degree to which an innovation adapts to the values, experiences, and the needs of the adopters

Perceived security

Degree to which ICT are trusted with the information they manage

Lifestyle

Level of ICT interaction within the community

In order to continue with ISM development, the ten ISM matrices obtained from the exercise described above were averaged, cell by cell, and the value ‘1’ was left in the cell, provided that the value of the average was seven or higher. As contemplated by the ISM methodology, at this point, the accessibility matrix was obtained, by weighing all of the consolidated responses. 3.4 Relationship Graph Based on the consensus of variables performed with expert validation and in an analysis of centralities (power of dependence and drive), the identified variable relationships and relevance are established (Fig. 2). For the construction of the degree, use was made of the canonical matrices resulting from the process with the ISM methodology and the Gephi software was used for its graphic representation. The construction took into account the matrices of the experts and the power of dependency and driving power.

154

D. M. Agudelo et al. Table 4. Expert Consultants

Expert

Description

1

Expert in innovation and social technology

2

Expert in appropriation, ICT, digital inclusion, and the digital gap

3

Expert In design, planning, management, and implementation of projects with high levels of use of ICT components and appropriation

4

Expert In design and evaluation of political publications related to the ICT sector and the society of information

5

Expert in social innovation appropriation and encouragement

6

Expert in innovation diffusion

7

Expert in digital citizenship

8

Expert in public employment and implementation of technological solutions for users

9

Expert in evaluation of web tool appropriation by media education students

10

Expert in digital ICT inclusion, adoption, and appropriation

11

Expert in e-learning impact evaluation

The intensity and diameter of the variable circles indicate the relevance of these within the constructed graph, as well as an analysis that considers their dependence power (degree of input) and their power of drive (degree of output). Based on this analysis, the most significant variables in ICT adoption and appropriation processes are the following: level of education, motivation, occupation, significant use of ICT products, and minors at home.

4 Conclusions The development of this model utilizes the ISM methodology, which concludes that race, ethnicity, and religion variables do not now, nor have they affected the other variables of consensus. This is due to the fact that their power of dependence and drive is 0. Conversely, the power of drive of confidence in the community and age of the population is 0, which means that these variables are affected by others, but do not affect any variables themselves (it is proposed that the most significant variables in the model are those that have low dependence power and high drive power). The most significant variables in the model highlighted in the conical matrix and based on their power of dependence and drive are: age, level of education, occupation, intellectual capacity (individual), and institutional support (group). The most significant variables identified on the graph are as follows: level of education, motivation, occupation, significant use of ICT products, and number of minors at home (individuals).

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation

155

Fig. 2. Expert consensus on variables for ICT adoption

It was also identified that the gap in digital inclusion and phenomena are multivariable, and that relationships between the variables that represent them are interdefinable (the study of any one of them requires study the others). As identified by the ISM methodology, and as mentioned by various authors in the literature reviewed, these phenomena are complex. In the literature review, which sought to encounter personal and group characteristics, the digital gap was found to not only reflect the difference in technological access but was also found to be the result of other social inequities. The expert consultation permitted a contrast of that which was identified in the bibliographic material, with their interpretation of ICT adoption and appropriation. Considering their perceptions and knowledge at the beginning of the process, ISM made it possible to eliminate biases.

156

D. M. Agudelo et al.

References 1. Becerra, G.: Sociocybernetics: tensions between complex systems, social systems and complexity sciences. Athenea Dig. 16(3), 81–104 (2016). https://doi.org/10.5565/rev/athenea. 1636 2. Schwarz, A., Chin, W.: Looking forward: toward an understanding of the nature and definition of IT acceptance. J. Assoc. Inf. Syst. 8(4), 230–243 (2007) 3. Mitrovic, Z., Klaas, N., Mbhele, F.: Benefits of introducing the cloud computing based mgovernment in the western cape: policy implications. In: ACM International Conference Proceeding Series, pp. 317–325. Association for Computing Machinery, Seoul (2013). https:// doi.org/10.1145/2591888.2591945 4. Surman, M., Reilly, K.: Appropriating the internet for social change: towards the strategic use of networked technologies by transnational civil society organizations. Social Change (November) (2003) 5. Bar, F., Weber, M.S., Pisani, F.: Mobile technology appropriation in a distant mirror: baroquization, creolization, and cannibalism. New Media Soc. 18(4,SI), 617–636 (2016). https:// doi.org/10.1177/1461444816629474 6. Camacho Jiménez, K.: Internet : a tool for social change? Elements of a necessary discussion (2007) 7. Jayant, A., Azhar, M., Singh, P.: Interpretive Structural Modeling (ISM) approach: a state of the art literature review. Int. J. Res. Mech. Eng. Technol. 4(3), 15–21 (2014). ISSN 2249-5770, I.F.= 3.907 8. Attri, R., Dev, N., Sharma, V.: Interpretive Structural Modelling (ISM) approach: an overview. Res. J. Manag. Sci. 2, 3–8 (2013) 9. Sun, H., Zhang, P.: (2006). The role of moderating factors in user technology acceptance. Int. J. Human-Comput. Stud. 64(2), 53–78. https://doi.org/10.1016/j.ijhcs.2005.04.013 10. Noce, A.A., McKeown, L.: A new benchmark for Internet use: a logistic modeling of factors influencing Internet use in Canada, 2005. Govern. Inf. Q. 25(3), 462–476 (2008). https://doi. org/10.1016/j.giq.2007.04.006 11. Maldifassi, J.O., Canessa, E.C.: Information technology in Chile: how perceptions and use are related to age, gender, and social class. Technol. Soc. 31(3), 273–286 (2009). https://doi. org/10.1016/j.techsoc.2009.03.006 12. Billon, M., Marco, R., Lera-Lopez, F.: Disparities in {ICT} adoption: a multidimensional approach to study the cross-country digital divide. Telecommun. Policy 33(10–11), 596–610 (2009). https://doi.org/10.1016/j.telpol.2009.08.006 13. Schleife, K.: What really matters: regional versus individual determinants of the digital divide in Germany. Res. Policy 39(1), 173–185 (2010). https://doi.org/10.1016/j.respol.2009.11.003 14. Weber, D.M., Kauffman, R.J.: What drives global {ICT} adoption? Analysis and research directions. Electron. Commer. Res. Appl. 10(6), 683–701 (2011). https://doi.org/10.1016/j. elerap.2011.01.001 15. Almuwil, A., Weerakkody, V., El-Haddadeh, R.: A conceptual study of the factors influencing e-inclusion. In: Proceedings of the European, Mediterranean and Middle Eastern Conference on Information Systems—Informing Responsible Management: Sustainability in Emerging Economies, EMCIS 2011, pp. 198–209. conference, Athens (2011) 16. González Zabala, M.P., Galvis Lista, E.A., Sánchez Torres, J.M.: Identificación de factores que afectan el desarrollo de la inclusión digital. Revista Virtual Universidad Católica Del Norte 1(44), 175–191 (2015) 17. Kubiatko, M.: The comparison of different age groups on the attitudes toward and the use of ICT. KURAM VE UYGULAMADA EGITIM BILIMLERI 13(2), 1263–1272 (2013)

Consensus of Individual and Group Characteristics for ICT Adoption and Appropriation

157

18. Taipale, S.: The use of e-government services and the Internet: the role of socio-demographic, economic and geographical predictors. Telecommun. Policy 37(4–5), 413–422 (2013). https:// doi.org/10.1016/j.telpol.2012.05.005 19. Pick, J.B., Sarkar, A., Johnson, J.: United States digital divide: state level analysis of spatial clustering and multivariate determinants of ICT utilization. Socioecon. Plann. Sci. 49, 16–32 (2015). https://doi.org/10.1016/j.seps.2014.09.001 20. Hinostroza, J., Matamala, C., Labbe, C., Claro, M., Cabello, T.: Factors (not) affecting what students do with computers and internet at home. Learn. Media Technol. 40(1), 43–63 (2015). https://doi.org/10.1080/17439884.2014.883407 21. Papaioannou, P., Charalambous, K.: Principals’ attitudes towards ICT and their perceptions about the factors that facilitate or inhibit ICT integration in primary schools of Cyprus. J. Inf. Technol. Educ.:Res. 10(1), 349–369 (2011) 22. Koç, T., Turan, A.H., Okursoy, A.: Acceptance and usage of a mobile information system in higher education: an empirical study with structural equation modeling. Int. J. Manag. Educ. 14(3), 286–300. https://doi.org/10.1016/j.ijme.2016.06.001 23. Thill, M.C., Rosenzweig, J.W., Wallis, L.C.: The relationship between student demographics and student engagement with online library instruction modules. Evid. Based Libr. Inf. Pract. 11(3), 4–15 (2016). https://doi.org/10.18438/B8992D 24. Nilashi, M., Ahmadi, H., Ahani, A., Ravangard, R., Bin Ibrahim, O.: Determining the importance of hospital information system adoption factors using fuzzy Analytic Network Process (ANP). Technol Forecast. Soc. Change 111, 244–264 (2016). https://doi.org/10.1016/j.tec hfore.2016.07.008

Trends of Artificial Intelligence and Blockchain in New Public Management Jhon Wilder Sanchez-Obando1(B)

and Luis Fernando Castillo Ossa2,3

1 Universidad de Manizales, Caldas, Colombia

[email protected]

2 Grupo de Investigación en Inteligencia Artificial, Universidad de Caldas, Manizales-Caldas,

Colombia [email protected] 3 Departamento de Ingeniería Industrial, Universidad Nacional de Colombia Sede Manizales, Manizales, Colombia

Abstract. Artificial Intelligence and Blockchain are technologies that are attracting more and more attention from different actors in various fields, including the New Public Management thanks to its unique features, such as decentralization, reliability, predictability, network distribution, profiling and security. Therefore, a study on the state of the new public management is considered necessary. The aim of the present study is to provide an overview of the state of the art related to the application of artificial intelligence and blockchain in the new public management to serve as a reference for future research in this field. For this purpose, a systematic literature review was conducted through the PRISMA protocol combined with the Tree of Science (ToS). In total, 2100 articles published between 2016 and 2023 in the Web of Science, Google Scholar, IEEE Xplore and Scopus databases participated in the study, where the Tree of Science (ToS) algorithm was subsequently applied to identify the most relevant literature on the topic. As a result, 81 papers were analyzed. The results indicate as research trends on the topic: supply chain, technology, digital business and technology adoption. Keywords: Artificial intelligence · Blockchain · New public management · SLR PRISMA · ToS

1 Introduction The blockchain as a technology has been of growing interest to the scientific community, however, as mentioned by [1], there is a procedure for research in blockchain, which establishes conceptual, descriptive and predictive levels. Therefore, it is necessary to investigate the contributions of knowledge to the different levels of research in blockchain with which it is possible to support the research problem from the informal and formal logic for the understanding of the reality of the phenomenon under study. Nakamoto [2], was the first person to introduce a cryptocurrency which he called: Bitcoin, and with this the blockchain technology. This made it possible to decentralize © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 158–169, 2023. https://doi.org/10.1007/978-3-031-36957-5_14

Trends of Artificial Intelligence and Blockchain in New Public Management

159

person-to-person economic transactions without the intermediation of a third party such as banks. To add a new block to the existing blockchain, it is necessary to record transactions and solve a cryptographic equation with high computational demand to reach consensus in the blockchain. From there, various platforms for blockchain development have emerged, such as Ethereum [3] and Hyper ledger Fabric [4]. Which unlike Bitcoin can work with smart contracts, which emerge as a proposal to the solution of problems in different sectors such as education [5]. Smart contracts are self-executing codes on blockchain to execute peer-to-peer agreement actions without any supervision [3]. The academic community focuses its research activities on different technologies of the fourth industrial revolution such as: blockchain, artificial intelligence, cloud computing and the internet of things. Which have a positive impact on the digital transformation of organizations [6]. Tapscott and Euchner [7], manifests, combined blockchain and artificial intelligence conceive the internet of value and efficiency that encourage profound changes in society and business. Some studies identify adoption issues such as lack of regulation, security and privacy, insufficient and lack of interoperable infrastructure, inefficient and costly transactions from an energy and transactional point of view, the need for value-oriented transitions in administrative processes and the absence of effective governance models [8, 9] and [10]. Other studies have identified governance as a key challenge for blockchain implementation in the public sector [11, 12] and [9] some propose to explore the implications of blockchain for public governance [11, 13–15] and [10]. However, blockchain governance continues to be a challenge for public sector organizations and instruments and models are needed to address the design, implementation, and monitoring of blockchain systems in public management [8] and [9]. This research involves determining how blockchain and artificial intelligence will affect new public management, particularly in AI-enabled decision-making assistance, and identifying issues that shape how blockchain and artificial intelligence will affect public service. This research reviews the literature on the applications of blockchain and artificial intelligence related to the new public management, thus contributing to it in three main ways. First, the literature review provides four categories of blockchain and artificial intelligence usage (supply chain; IT technology; digital business and technology adoption). Secondly, the use of hybrid scientometric techniques Tree of Science (ToS) and bibliometrix to improve the results under the PRISMA protocol and thirdly and thirdly the use of blockchain technologies and artificial intelligence as models and Tools for the development of the new public management. This article is organized as follows. Section 2 presents the PRISMA protocol review procedure adopted in this study. Section 3 shows the results of the systematic literature review. Section 4 provides an in-depth analysis and discussion of the results obtained. Based on the results of the systematic literature review, Sect. 5 identifies knowledge gaps. Section 5 presents the conclusions of the study.

160

J. W. Sanchez-Obando and L. F. Castillo Ossa

2 Research Methodology For this reason, it is proposed in this study to propose a SLR PRISMA with software tools to complement the systematic literature review to find the trends of blockchain and artificial intelligence and new public management. The search methodology proposed by [16], allows for a systematic literature review that explores the relationship between blockchain and artificial intelligence and new public management. 2.1 Definition of Research Question This step consists of establishing the questions guiding the research in order to obtain a detailed vision of the topic addressed. The objectives of the research are (i) to discover blockchain and artificial intelligence and new public management trends (ii) to classify blockchain and artificial intelligence and new public management research (iii) to identify blockchain and artificial intelligence and new public management research cooperation networks and (iv) to analyze blockchain and artificial intelligence and new public management trends research opportunities. The research questions are as follows: RQ1: ¿ What is the current state of play on scenarios for the use of blockchain and artificial intelligence in new public management? RQ2: ¿ What is the behavior of the authors who publish on the topic of blockchain and artificial intelligence and new public management? RQ3: ¿ What are the characteristics of blockchain and artificial intelligence that could benefit the new public management? RQ4: ¿What are the research trends of blockchain and artificial intelligence and new public management? You can be seen that 4 questions were asked. Objective 1 has assigned SLR questions RQ1 and RQ2 related and objective 2 has RQ3 and RQ4 related. 2.2 Search In [18], develops a procedure for the construction of a search equation, which starts with the search for SLRs addressed in the research topic blockchain and artificial intelligence where the following results were obtained 8.33% a literature review, 11.1% a systematic literature review and 12.5% a review of studies. From the above results, 15 SLRs were found in the topic and set as the search equation: Artificial and blockchain and new public management. Once this stage is finished, we proceed to write the justified search equation in each of the databases: Scopus, Web of Science (WoS), Google Scholar and IEEE Xplore. The search was performed in January 2023. To search the articles, we used the search equation: blockchain and artificial intelligence and new public management. Table 1 presents the structure of the search equations for each database. Based on the table above, it can be seen that the database that requires the greatest complexity in the construction of the search equation is the Scopus database. At the end of this stage 81 articles, as shown in Fig. 1. Figure 1 shows the classification procedure of the articles found to which exclusion criteria are applied protocol PRISMA.

Trends of Artificial Intelligence and Blockchain in New Public Management

161

Table 1. Searches in database Scientific database

Query strings

Scopus

“Artificial and blockchain and new public management”

Web of Science (WoS)

“Artificial and blockchain and new public management”

Google Scholar IEEE Xplore

“Artificial intelligence and blockchain and new public management” “Artificial and blockchain and new public management”

Fig. 1. Publication founds. Source: Own elaboration based on [19].

The Fig. 1, it can be seen that the number of studies with which the PRISMA protocol for the construction of the SLR was performed is 81 once the exclusion criteria were established in the search equation. To find the articles in the Google Scholar database, it was necessary to use the Publish or Perish (PoP) software to extract the documents by items: title, abstract and authors. It is important to mention that the number of documents excluded by other criteria in the PRISMA protocol is due to the exclusion processes in the databases in which documents developed in other areas of knowledge such as: philosophy, psychology, mathematics, physics, chemistry were excluded: chemical engineering, cosmology, decision sciences and economics. You be seen that the research topic of artificial intelligence and blockchains in the new public management has been the one of greatest interest for the academic community since 2019, 2022. Although the first works were developed in 2016, the production it is very low compared to the years 2019 and 2022. The graph above represents the annual production of the four databases: Google Scholar, IEEE Xplore, Scopus and WoS.

162

J. W. Sanchez-Obando and L. F. Castillo Ossa

2.3 Study Selection This step consists of choosing the relevant papers that would help answer the research questions. To do so, a set of inclusion and exclusion criteria are defined in Table 2. Table 2. Inclusion and exclusion criteria for relevant documents Inclusion criteria

Exclusion criteria

Documents demonstrating the application of blockchain in education that are part of the grey literature

Editorial Comments and Opinions

Papers proposing feasible solutions to blockchain problems in education (method, technique, model and conceptual framework)

Documents that present surveys

Papers proposing solutions that have been evaluated (implemented, simulated and mathematically modeled) that are part of the gray literature

Documents written in languages other than English

Documents produced in English only Papers published in journals and conferences that are part of the gray literature

Books

Based on the above results, we proceeded to exclude documents outside the scope of this study. Thus, only documents that met all the inclusion criteria were included, filtering the studies according to the title, abstract and list of keywords. 2.4 Data Extraction This stage consists of gathering the information necessary to address the research questions of this study. Therefore, we established the review criteria that contained 9 review criteria, implemented in 2100 documents included in this scientometric review: name authors, title, DOI, affiliation, abstract, keywords, type documents, language and source.

3 Results We proceed to analyze the 81 articles through a network analysis in the construction of the science tree supported by the Bibliometrix platform to answer each of the research questions. The initial search on the search equation defined in the systematic literature review was carried out in Web of Science, Scopus, IEEE Xplore and Google Scholar with the use of Publish or Perish (PoP). The results were exported to the ToS scientometric platform and Bibliometrix plataform that generates a tree structure, in its root are located the classic articles, in the trunk the structural documents and in the leaves the recent documents, in this way subcategories are determined by citation analysis.

Trends of Artificial Intelligence and Blockchain in New Public Management

163

The first 81 documents of the search result are exported to a file in txt format, and so on until 81 documents are found, which will be the seed of the tree of science for the bibliographic analysis process, using an algorithm based on graphs. The data in the seed correspond to the name of the authors, name of the journal, journal edition, abstract, keywords, affiliation, country of origin, DOI and the references cited in each article; these data allow classifying the results to establish a category analogous to one of the parts of a tree (root, stem and leaves). The.txt file is loaded into the R Studio Cloud software, a new project is generated and the ToS algorithm is installed, which generates a graph that is composed of nodes and edges, in which the articles or research papers are the nodes and the connections between them are the edges. And articles that cite others in the network (root and trunk) but are not cited, will be leaves [19]. The perspectives of blockchain and artificial intelligence and new public management, are constructed by citation analysis, through a clustering algorithm [20], then the concepts that compose each perspective are identified and analyzed from text mining of the R Cloud package. The ToS algorithm is complemented by the bibliometrix development by [22] tool that focuses on the visualization of the behavior of the data extracted from the consulted databases. 3.1 Publication and Geographical Distribution The distribution of the selected articles by publication year is presented in Fig. 2 for each database.

Fig. 2. Annual Production for Countries in Scopus and WoS. Source: Own elaboration based in Bibliometrix.

The Fig. 2, it is observed that for the Scopus and WoS databases there are differences in the production by countries in the world. For the Scopus database, 45 countries that make up the academic community are observed, of these countries, China, USA, India, Australian, Latinamerican and Europe stands out with a darker color that reflects the

164

J. W. Sanchez-Obando and L. F. Castillo Ossa

intensity of production, therefore, China, USA, India, Australian, Latinamerican and Europe are the world zones with the highest academic production for thematic blockchain and artificial intelligence and new public management. On the other hand, the WoS database shows 14 countries, which includes Latin American countries such as Brazil, it also highlights those countries such as China and Australian are leaders with darker color that reflects the intensity of production of these countries worldwide as leaders in the following areas blockchain and artificial intelligence and new public management. 3.2 Summary of Contributions Figure 3 shows the cluster of ToS scientometric in the Scopus and WoS databases.

Fig. 3. Cluster of trends in Scopus and WoS. Source: Own elaboration based in ToS scientometric.

Figure 3, the trends found as a result of the ToS algorithm can be observed, therefore, we proceed to describe each of the four clusters found. – Trend 1: In this cluster are the documents related to the identification of supply chain in new public management, in which the root of the tree in this cluster is [23], presents an algorithm for sequencing in industry 4.0 or smart industries. On the other hand, the trunk considered as a structural document of the solutions proposed by blockchain and artificial intelligence in new public management is [24], develops a study to analyze the strength of correlation between the use of big data and function optimization in operations research in organizations. And as a leaf we have [25], they propose a multi-criteria decision making model to address management risks in supply chains. – Trend 2: This includes documents related to the development of Technology computer, with a focus on models build in which the root of the tree of this cluster is [26], proposes a network model to improve information transfer problems in smart contracts. On the other hand, the trunk considered as a structural document in the development of blockchain technology is [27], who through an exploratory study evaluate the concept of Industry 4.0 and its effects on the factors that determine

Trends of Artificial Intelligence and Blockchain in New Public Management

165

the implementation of the circular economy in a country. And as a leave have [28], who proposed a model of digital transformation in organizations where the impact of blockchain technology on organizational management is evaluated. – Trend 3: In this are the documents related to digital business, in which the root of this cluster is [29], have proposed a network of a micropayment channel for the transfer of bitcoins. On the other hand, the trunk considered as a structural document in the digital business with blockchain and artificial intelligence is [30], who proposed a multi-agent system to improve consensus routing of bitcoin transactions in digital businesses. [31], who proposed smart contracts to facilitate multi-user collaborative control to facilitate the validation of documents transferred on the blockchain. – Trend 4: In this are the documents related to technology adoption, in which the root of this cluster is [32], who proposed a multiheuristic technique for rapid decision making with a focus on decision makers in the public sector. On the other hand, the trunk considered as a structural document in the technology adoption with blockchain and artificial intelligence is [33], who developed an economic analysis of blockchain transaction costs that provides a blockchain infrastructure for organizations. [34], who establish a regulatory framework and monitor the proliferation of digital businesses financed by initial cryptocurrency offerings. 3.3 Contributions by Authors and Characteristics for Blockchain and Artificial Intelligence and New Public Management In this section we want to present the main contributions of a similar SLR with the research topic presented in this article. Figure 4 shows the thematic map of blockchain and artificial intelligence research.

Fig. 4. Artificial intelligence system in relation blockchain and taxonomy of blockchain for artificial intelligence. Source: [34].

Based on contribution of [35], specific applications in the areas of agriculture, medicine and finance, indicating the advantage of the hybridization of artificial intelligence and blockchain technologies in complex systems such as public management.

166

J. W. Sanchez-Obando and L. F. Castillo Ossa

4 Discussion The systematic review of the 81 publications mentioned above provided the information necessary to answer the three initial research questions. RQ1: ¿ What is the current state of play on scenarios for the use of blockchain and artificial intelligence in new public management? Although there is a high number of documents extracted from the databases 2100 documents by the PRISMA protocol illustrated in Fig. 1 is reduced to only 81 documents of interest that based on Figs. 1 and 2 it is observed that it is an area or topic of research that increasingly attracts the attention of the academic community. Based on Fig. 4, the academic production for Latin America is concentrated in Brazil and North America, with the United States as leader, and for Europe and Australian, China. The Scopus database indicates that research is concentrated in fields such as: industry 4.0, artificial intelligence, blockchain and digital transformation. RQ2: ¿ What is the behavior of the authors who publish on the topic of blockchain and artificial intelligence and new public management? The authors in the trends showed that there is a concern in overcoming the problems and technological gaps of the blockchain to be implemented in different sectors for this reason, there is a trend of authors for the year 2022 proclaiming as research problems the distribution, modeling networks and graphs. From the SLRs proposed by [35], it is observed that few articles related to the object of the present study, for the case of blockchain and artificial intelligence and new public management 81 articles were evidenced in the SLR of Salah where there is a greater preference by the academic community to investigate in artificial intelligence and blockchain in topics: finance, energy and services business. RQ3: ¿ What are the characteristics of blockchain and artificial intelligence that could benefit the new public management? According to [36], blockchain has ten characteristics (distributed consensus, transaction verification, platforms for smart contracts, peer-to-peer value transfer, cryptocurrency/incentive generation, smart property, security provision, immutability, uniqueness, and smart contracts) that they make it worthy of investigation to contribute to the new public management. [37], proposed states that decentralization through public or private networks facilitates the implementation of technology at the following levels: Smart Energy, Smart Cities, Smart Cities and Sharing Economy, Smart Government, Smart Homes, Smart Transportation and Construction Management and Business Models and Organizational Structures combined in a four-dimensional model supported by a banking decision-making model. RQ4: ¿What are the research trends of blockchain and artificial intelligence and new public management? To answer this SLR question, the ToS algorithm was applied to find the trends and the number of articles per trend, classifying the oldest articles as corresponding to the root of the tree. The reference or classic articles, the most modern articles correspond to the trunk of the tree which correspond to the structural articles. They cite the root and are cited by future studies, and the most modern articles are the leaves of the tree which are characterized by citing the studies of the trunk of the tree and the root. For the present SLR, few articles were found and in total the algorithm applied allowed us to find 4 trends that are described in Fig. 3.

Trends of Artificial Intelligence and Blockchain in New Public Management

167

5 Conclusions The result of this research is expressed through the analogy of a tree, the co-occurrence of keywords and collaboration networks between countries, so that in this way the evolution of this area of knowledge is explicitly understood. The root jobs were considered as the base or starting point of the relationship between blockchain and budget allocation and public education, the trunk jobs provide the structure of the concept, and the leaf jobs are considered as the blockchain trend jobs in budget allocation. And public education. Blockchain and artificial intelligence can greatly contribute to decision-making, e-government, decentralization, security, and traceability in the new public management, which is evident in the different initiatives that have been investigated and published in recent years. The trends shown in supply chains, technology, digital business, and technology adoption show that there are research opportunities from other fields of knowledge such as new public management. Some studies reflected by the science tree indicate that the knowledge gaps lie in the scalability of blockchain technology in different fields of knowledge and the technical difficulties of adopting blockchain technology in public administration. For future research, it is recommended to use the Perish or Public (PoP) software in addition to the tools mentioned above, to rely on the results of google scholar in the systematic literature review.

References 1. Zhao, S., Li, S., Yao, Y.: Blockchain enabled industrial Internet of Things technology. IEEE Trans. Comput. Soc. Syst. 6(6), 1442–1453 (2019) 2. Nakamoto, S.: Bitcoin: a peer-to-peer electronic cash system. Decentralized Bus. Rev. 21260 (2008) 3. Buterin, V.: A next-generation smart contract and decentralized application platform. White Pap. 3(37), 1–2 (2014) 4. Androulaki, E., et al.: Hyperledger fabric: a distributed operating system for permissioned blockchains. In: Proceedings of the Thirteenth EuroSys Conference, pp. 1–15 (2018) 5. Yasin, A., Liu, L.: An online identity and smart contract management system. In: 2016 IEEE 40th Annual Computer Software and Applications Conference (COMPSAC), vol. 2, pp. 192– 198 (2016) 6. Benlian, A., Kettinger, W.J., Sunyaev, A., Winkler, T.J., Editors, G.: The transformative value of cloud computing: a decoupling, platformization, and recombination theoretical framework. J. Manag. Inf. Syst. 35(3), 719–739 (2018) 7. Tapscott, D., Euchner, J.: Blockchain and the internet of value: an interview with Don Tapscott Don Tapscott talks with Jim Euchner about blockchain, the Internet of value, and the next internet revolution. Res. Manag. 62(1), 12–19 (2019) 8. Janssen, M., Weerakkody, V., Ismagilova, E., Sivarajah, U., Irani, Z.: A framework for analysing blockchain technology adoption: integrating institutional, market and technical factors. Int. J. Inf. Manag. 50, 302–309 (2020). https://doi.org/10.1016/j.ijinfomgt.2019. 08.012 9. Ølnes, S., Ubacht, J., Janssen, M.: Blockchain in government: benefits and implications of distributed ledger technology for information sharing. Govern. Inf. Q. 34(3), 355–364 (2017). Elsevier

168

J. W. Sanchez-Obando and L. F. Castillo Ossa

10. Zachariadis, M., Hileman, G., Scott, S.V.: Governance and control in distributed ledgers: understanding the challenges facing blockchain technology in financial services. Inf. Organ. 29(2), 105–117 (2019) 11. Atzori, L., Iera, A., Morabito, G.: Understanding the Internet of Things: definition, potentials, and societal role of a fast evolving paradigm. Ad Hoc Netw. 56, 122–140 (2017) 12. Meijer, D., Ubacht, J.: The governance of blockchain systems from an institutional perspective, a matter of trust or control? In: Proceedings of the 19th Annual International Conference on Digital Government Research: Governance in the Data age, pp. 1–9 (2018) 13. Beck, R., Müller-Bloch, C., King, J.: Governance in the blockchain economy: a framework and research agenda. J. Assoc. Inf. Syst. 19(10) (2018). https://aisel.aisnet.org/jais/vol19/iss 10/1. Accessed 05 Nov 2021 14. De Filippi, P., Mannan, M., Reijers, W.: Blockchain as a confidence machine: the problem of trust & challenges of governance. Technol. Soc. 62 (2020). https://doi.org/10.1016/j.techsoc. 2020.101284 15. Werbach, K.: The Blockchain and the New Architecture of Trust. Mit Press (2018) 16. Yepes-Nuñez, J.J., Urrútia, G., Romero-García, M., Alonso-Fernández, S.: The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Rev. Esp. Cardiol. 74(9), 790–799 (2021). https://doi.org/10.1016/j.recesp.2021.06.016 17. Napoleao, B.M., et al.: Establishing a search string to detect secondary studies in software engineering. In: Proceedings of 2021 47th Euromicro Conference Software Engineering Advanced Applications SEAA 2021, pp. 9–16 (2021). https://doi.org/10.1109/SEAA53835. 2021.00010 18. Haddaway, N.R., Page, M.J., Pritchard, C.C., McGuinness, L.A.: PRISMA2020: an R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and open synthesis. Campbell Syst. Rev. 18(2), e1230 (2022). https://doi.org/10.1002/cl2.1230 19. Zuluaga, M., Arbelaez-Echeverri, O., Robledo, S., Osorio-Zuluaga, G.A., Duque-Méndez, N.: Tree of Science—ToS: a web-based tool for scientific literature recommendation. Search Less, Research More!, Issues Sci. Technol. Librariansh. no. 100, August (2022). https://doi. org/10.29173/ISTL2696 20. Blondel, V.D., Guillaume, J.L., Hendrickx, J.M., De Kerchove, C., Lambiotte, R.: Local leaders in random networks. Phys. Rev. E—Stat. Nonlinear, Soft Matter Phys. (2008). https:// doi.org/10.1103/PhysRevE.77.036114 21. Aria, M., Cuccurullo, C.: Bibliometrix: an R-tool for comprehensive science mapping analysis. J. Informetr. 11(4), 959–975 (2017) 22. Ivanov, D., Dolgui, A., Sokolov, B., Werner, F., Ivanova, M.: A dynamic model and an algorithm for short-term supply chain scheduling in the smart factory industry 4.0. Int. J. Prod. Res. 54(2), 386–402 (2016) 23. Choi, T., Wallace, S.W., Wang, Y.: Big data analytics in operations management. Prod. Oper. Manag. 27(10), 1868–1883 (2018) 24. da Silva, E.M., Ramos, M.O., Alexander, A., Jabbour, C.J.C.: A systematic review of empirical and normative decision analysis of sustainability-related supplier risk management. J. Clean. Prod. 244, 118808 (2020) 25. Alabi, K.: Digital blockchain networks appear to be following Metcalfe’s law. Electron. Commer. Res. Appl. 24, 23–29 (2017) 26. Pham, T.T., et al.: Industry 4.0 to accelerate the circular economy: a case study of electric scooter sharing. Sustainability 11(23), 6661 (2019) 27. Rehman Khan, S.A., Yu, Z., Sarwat, S., Godil, D.I., Amin, S., Shujaat, S.: The role of block chain technology in circular economy practices to improve organisational performance. Int. J. Logist. Res. Appl. 25(4–5), 605–622 (2022)

Trends of Artificial Intelligence and Blockchain in New Public Management

169

28. Poon, J., Dryja, T.: The bitcoin lightning network: scalable off-chain instant payments. (2016) 29. Korinek, A., Stiglitz, J.E.: Artificial intelligence and its implications for income distribution and unemployment. In: The Economics of Artificial Intelligence: An Agenda. University of Chicago Press, pp. 349–390 (2018) 30. Nizamuddin, N., Salah, K., Azad, M.A., Arshad, J., Rehman, M.H.: Decentralized document version control using ethereum blockchain and IPFS. Comput. Electr. Eng. 76, 183–197 (2019) 31. Cairney, P., Kwiatkowski, R.: How to communicate effectively with policymakers: combine insights from psychology and policy studies. Palgrave Commun. 3(1), 1–8 (2017) 32. Berg, C., Davidson, S., Potts, J.: Blockchain technology as economic infrastructure: revisiting the electronic markets hypothesis. Front. Blockchain 2, 22 (2019) 33. Kenney, M., Zysman, J.: Unicorns, Cheshire cats, and the new dilemmas of entrepreneurial finance. Ventur. Cap. 21(1), 35–50 (2019) 34. Salah, K., Rehman, M.H.U., Nizamuddin, N., Al-Fuqaha, A.: Blockchain for AI: review and open research challenges. IEEE Access 7, 10127–10149 (2019) 35. Bashir, I.: Mastering Blockchain: Deeper Insights into Decentralization, Cryptography, Bitcoin, and Popular Blockchain Frameworks. Packt Publishing Limited (2017) 36. Liu, X., Tang, Z., Li, P., Guo, S., Fan, X., Zhang, J.: A graph learning based approach for identity inference in dapp platform blockchain. IEEE Trans. Emerg. Top. Comput. (2020)

A Novel Approach for Smart Mobility Roadmap in Smart Cities Tatiana Zona-Ortiz

and Gustavo Guzman(B)

Universidad Santo Tomás. Sede Principal, División de Ingenierías, Facultad de Ingeniería de Telecomunicaciones, grupo INVTEL, Carrera 9 No. 51-11, Bogotá D.C., Colombia [email protected], [email protected]

Abstract. A novel approach for Smart Mobility Roadmap in Smart Cities is presented. The documents related to Key Performance Indicators (KPI) in smart cities of three internationals organizations was studied, and the evolution from an existing city to a Smart and sustainable city was considering. The main result is a set of 25 KPIs for smart mobility based on a comparison between KPIs proposed by International Telecommunications Union (ITU), International Organization for Standardization (ISO) and Colombia ITC Ministry (MINTIC). It is important to highlight that the roadmap highly depends on the city budget, the number of KPIs considered, policy priorities, and people adoption. Keywords: Smart mobility · Smart city · Information communication technology ‘ITC’ · Smart city KPIs

1 Smart Cities Overview In 2015, The United Nations (UN) built a plan with 17 objectives for removing poverty, improving human lives, and protecting the planet. The eleventh objective is about Sustainable Cities and Communities, this objective was thought-out due to the growth in cities and overloading basic services for a living. Therefore it is expected by 2050 80% of the people will be living in cities [1], contributing to 70% of CO2 emissions and consuming 60% of vital resources [2]. According Ref. [3], a group of experts from different fields is needed to build and design a Smart City, and they must consider fields like economic sciences, sociology, engineering, political science, and Information and Communication Technologies (ICT). Nonetheless, many frameworks have been proposed to structure Smart Cities, and the model proposed by the National Institute of Standards and Technology (NIST) is one of the most used. This model defines a Smart City as a complex system called a ‘System of Systems, and it usually is based on six components: government, mobility, economy, environment, living, and people. In the same way Ref. [3] declares that a Smart City is related to the systematic use of digital technologies by its people to achieve its main objectives to improve quality of life and sustainability. These cities are correlated with the intelligent use of infrastructure, energy, living, mobility, services, and security solutions, most of which are based on © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 170–182, 2023. https://doi.org/10.1007/978-3-031-36957-5_15

A Novel Approach for Smart Mobility Roadmap in Smart Cities

171

technologies such as blockchain, digital shadows, and IoT among others. These technologies should be accessible, structured, scalable, and impartial and the cities could adopt them by developing six different areas: smart environment, smart living, smart economy, smart mobility, smart government, and smart people. It means that this is correlated with NIST components. The International Telecommunications Union (ITU) also proposes a Framework of Smart and Sustainable City (SSC) standards [4] which is compounded by four layers: Buildings and physical infrastructure, Information and communication technology, SSC services, and Smart City management and assessment. Based on this framework, Ref. [1] propose two additional layers related to governance and strategy. As well, The International Organization for Standardization (ISO) proposes five documents related to Smart Cities Framework [5–9]. Although there has been much talk about frameworks, be it from the point of view of components, layers, or structure. The smart city implementation has to face many challenges such as budget, the city’s physical infrastructure, the adoption of technology by its inhabitants, and the availability of data in existing information systems. It is possible to face these challenges through a roadmap that makes it possible to visualize the general panorama in the implementation of a smart city. This implementation should be approached through verticals, one of the most important for the quality of life in a large city is mobility. 1.1 Smart Mobility as a Driver in the City The smart city implementation may improve the quality of life by optimizing the city services or improving the efficiency of urban systems, for example, the reduction of daily traveling time. Even though a smart city is not smart just for having many ICT facilities, but also to invest in citizens or respond to stakeholder requests, for instance, in transportation infrastructure (both modern and traditional), governance, and social capital [10]. Based on the previous statements and the phenomenon of urban growth in the world, mobility is a cross-axis in social and technological city planning. In this case, Smart Mobility is an interesting research field to improve transportation and urban development inside a Smart City. Smart Mobility is divided into six principal sub-areas according to [11]: Safe Driving, Intelligent Lighting Systems, Shared Urban Mobility, Electric Mobility or Transport, Green Mobility, and Intelligent Payment Systems. These subareas should be assessed using a standard method, for this reason, some regulatory entities have created some Key Performance Indicators (KPIs) to estimate its evolution or progress in a smart city according to city and people behavior. From the point of view of this study, the difference between smart mobility and smart living is mainly that the former affects only transportation or mobility issues, while the latter is much broader because it considers the entire city infrastructure that affects its inhabitants.

172

T. Zona-Ortiz and G. Guzman

2 Importance of Key Performance Indicators to Build Smart Mobility in Smart Cities KPIs are tools to manage any kind of project, they have been used to measure behavior, evolution, and results [12]. Therefore, KPIs play a main role in the process of building a Smart City [13], and, consequently, to build Smart Mobility. For this reason, The International Telecommunication Union (ITU) and the International Organization for Standardization (ISO) bring out some documents where a Smart City is modeled with some verticals or dimensions, and two of those verticals are highlighted in Smart Mobility: transportation and infrastructure [14, 15]. The following subsections focus on a selection of KPIs for Smart Mobility to lead the final roadmap approach. 2.1 Smart Mobility KPIs by the International Organization for Standardization Three documents of the International Organization of Standardization ISO were studied. Firstly, Sustainable cities and communities (ISO 37100), secondly the indicators for city services and quality of life (ISO37120), and finally the final draft of the International Standard of Sustainable Cities and Communities Indicators for Smart Cities ISO/FDIS 37122 [15] Fig. 1 shows how these documents are related.

Fig. 1. Relation between studied documents in ISO standards family. (Source ISO 37122)

Based on Smart City KPIs presented in ISO 37122, fifteen KPIs were extracted considering their high relationship with Smart Mobility. This selection represents 18.75% of KPIs presented in the document, so Table 1 shows the fourteen KPIs in the transportation vertical, and one of the ten that compound the energy vertical. These fifteen KPIs were prioritized because they encompass all the key aspects related to mobility as presented in the ISO 37122 document. The other KPIs outlined in the document are not directly relevant to improving the quality of life through the Smart Mobility field. Additionally, to facilitate KPIs comparison in the last column of Table 1 a unique ID was given for each extracted KPI. In these fifteen KPIs, clean transportation represents 33.3%, 26.6% is for traffic management, and another 26.6% is for high ICT applications. The remaining 13.3% is

A Novel Approach for Smart Mobility Roadmap in Smart Cities

173

Table 1. ISO KPIs applicable for Smart Mobility. (Source ISO 37122). **Authors Area

KPI

KPI SC ISO ID**

Transportation

Percentage of city streets and thoroughfares covered by SC-ISO-1 real-time online traffic alerts and information Number of users of sharing economy transportation

SC-ISO-2

Percentage of vehicles registered in the city that are low-emission vehicles

SC-ISO-3

Number of bicycles available through municipally provided bicycle-sharing services

SC-ISO-4

Percentage of public transport lines equipped with a publicly accessible real-time system

SC-ISO-5

Percentage of the city’s public transport services covered by a unified payment System

SC-ISO-6

Percentage of public parking spaces equipped with e-payment systems

SC-ISO-7

Percentage of public parking spaces equipped with real-time availability systems

SC-ISO-8

Percentage of traffic lights that are intelligent/smart

SC-ISO-9

City area mapped by real-time interactive street maps as SC-ISO-10 a percentage of the city’s total land area

Energy

Percentage of vehicles registered in the city that are autonomous vehicles

SC-ISO-11

Percentage of public transport routes with municipally provided and/or managed Internet connectivity for commuters

SC-ISO-12

Percentage of roads conforming with autonomous driving systems

SC-ISO-13

Percentage of the city’s bus fleet that is motor-driven

SC-ISO-14

Number of electric vehicles charging stations per registered electric vehicle

SC-ISO-15

allocated to traffic/mobility payment methods. It shows that ISO has a holistic vision of what means a Smart City. However, it can be observed that the KPIs introduce data related to the facilities that an ideal city should have soon. 2.2 Smart Mobility KPIs by International Telecommunication Union The International Telecommunication Union (ITU) has developed a set of KPIs to help cities measure their progress in achieving the Sustainable Development Goals (SDGs), becoming smarter and more sustainable [14]. This set of 91 KPIs is organized into three main dimensions: Society and Culture, Economy, and Environment. These dimensions

174

T. Zona-Ortiz and G. Guzman

use subdimensions and categories to group KPIs by topics, there are also two types of indicators: core and advanced [13]. KPIs attempt to collect data in a consistent and standardized way, so cities can measure their performance, compare their progress with other cities, and share best practices. The set of KPIs also serves as the basis for the initiative United for Smart Sustainable City Index (U4SSC), which collects reported KPI values and data about the city profile to provide a comparative ranking of cities. Each indicator is described, including the rationale for choosing the indicator, how it should be interpreted, what benchmarking trends are considered desirable, the methodology for calculating the value to be reported, and potential sources of data. The set of KPIs is a tool to provide cities with a means for self-assessment of the SDGs [14]. Table 2. Smart mobility KPIs according to ITU (Source: ITU). **Authors Dimension

Sub-dimension

Category

KPI

Economy

ICT

Transport

Dynamic public SC-ITU-1 transport information

Core

Traffic monitoring

SC-ITU-2

Core

Intersection control

SC-ITU-3

Advanced

Infrastructure

Society and culture

Safety, housing and social inclusion

Transport

Safety

KPI SC ITU ID**

TYPE

Public transport SC-ITU-4 network

Core

Public transport SC-ITU-5 network convenience

Advanced

Bicycle network SC-ITU-6

Core

Transportation mode share

SC-ITU-7

Advanced

Travel time index

SC-ITU-8

Advanced

Shared bicycles

SC-ITU-9

Advanced

Shared vehicles

SC-ITU-10

Advanced

Low-carbon emission passenger vehicles

SC-ITU-11

Advanced

Traffic fatalities

SC-ITU-12

Core

A Novel Approach for Smart Mobility Roadmap in Smart Cities

175

The set of KPIs has 91 indicators, of which 12 were selected concerning Smart Mobility. This selection considers all the KPIs in the transport category and one in the safety category, and it represents 13.18% of the KPIs proposed by ITU. The KPIs selection is shown in Table 2. In these KPIs selection, traffic management and traffic safety represent 41.7%, while clean transportation is 33.3%, and high ICT is 25%. As can be seen, the ITU’s KPIs were primarily designed for present cities to collect current data, but they also consider high ICT for cities that are ahead of others. 2.3 Smart Mobility KPIs in Colombia Colombia´s ICT Ministry (MINTIC Spanish acronym) proposes a model to measure Smart Cities and territories. This model involves six dimensions and twenty-nine subdimensions to organize 126 KPIs, it also has five enabling axis. Figure 2 shows people at the center of the six proposed dimensions since people are the agents of the four helixes of society [16]. Inside this model, the focal point for Smart Mobility is the Habitat dimension, because this dimension contains the Smart Mobility subdimension and the Smart Infrastructure subdimension [17]. All the KPIs in the Smart Mobility subdimension and three of the five KPIs in the Smart Infrastructure subdimension were selected. This selection represents 42.85% of KPIs in the habitat dimension and 9.52% of KPIs in the Colombian initiative.

Habitat Economy

People Society Industry Goverment - Academy Environment

Governance Quality of life

Fig. 2. Dimensions and society helixes of territories and Smart cities. MINTIC Colombia

The percentage breakdown for MINTIC’s KPIs is 58% for traffic management and traffic facilities and 41.7% for clean transportation. This proposal from MINTIC does not include the measurement of high ICT applications. MINTIC initiative has KPIs focused on measuring present cities and using current data, as well as ITU (Table 3).

176

T. Zona-Ortiz and G. Guzman

Table 3. Smart mobility KPIs according to MINTIC Colombia. (Source MINTIC COL) **Authors Dimension Sub-dimension

KPI

KPI SC COL ID**

Habitat

Clean and efficient public transport

SC-COL-1

Use of public transport per travels

SC-COL-2

Smart mobility

Use of public transport per occupation SC-COL-3 Dynamic public transport information SC-COL-4 Electric vehicle charging stations

SC-COL-5

Traffic rates per time waisted

SC-COL-6

Traffic rates per mean speed

SC-COL-7

Accident rate

SC-COL-8

Traffic fatalities Smart infrastructure Bike paths

SC-COL-9 SC-COL-10

Air connectivity

SC-COL-11

Intercity transportation

SC-COL-12

3 Smart Mobility KPIs Comparison Table 4 compares KPIs proposed for Smart Mobility by ITU, ISO, and MINTIC. This comparison uses KPI ID in the tables above to relate the different KPIs presented. It is important to highlight that how to measure each KPI was not indicated, however, the relation may be established through the KPI name and description. The measurement column in Table 4 is defined according to the measurement proposed by involved organizations and KPI interpretation. This measurement is based on the application in environments where technological advances are at a moderate level. This set of KPIs has twenty-five KPIs. Surprisingly, only KPI number 4 is coexistent in the three studied organizations, which allows great flexibility to establish a roadmap. There are fourteen KPIs without any coincidence between the studied organizations, which means 56% of the set. In detail, five KPIs belong to the MINTIC initiative, eight belong to ISO documents and one belongs to ITU recommendations. The main disadvantage is it increases the size of the set of KPIs. On the other hand, the main advantage is flexibility due to the diversity of KPIs, which can increase the time for executing a roadmap. Furthermore, there are cases where KPIs only coincide for two of the three studied organizations. Indeed, in this set of KPIs, four KPIs are in ITU and ISO, two are MINTIC and ISO and four are in MINTIC and ITU. KPIs in the MINTIC initiative are correlated with other organizations in 28% of the set of KPIs and they are unique for the MINTIC initiative for the 20% of the set.

SC-COL-6

SC-COL-7

SC-COL-8

SC-COL-9

SC-COL-10

SC-COL-11

SC-COL-12

6

7

8

9

10

11

12

SC-ISO-15

SC-ISO-1

SC-COL-5

5

SC-ISO-5

15

SC-COL-4

4

SC-ISO-3

SC-COL-3

3

14

SC-COL-2

2

SC-ISO-14

SC-ISO-4

SC-COL-1

1

KPI SC ISO ID

13

KPI SC COL ID

No

SC-ITU-2

SC-ITU-11

SC-ITU-6

SC-ITU-12

SC-ITU-8

SC-ITU-1

SC-ITU-5

KPI SC ITU ID

(continued)

Medium term

Medium term

Short term

#Lowemissionpassengervehicles ∗ 100 #Totalvehicles #kilometersoftrafficmonitoredways #Kilometersofallprincipalways ∗ 100

Short term

#Bicyclessharedavailable #10.000people

Short term

# Qty of transport urban lines

Short term

#flightroutes 10.000people

Short term

#trafficfatalities 10.000people

#Bike routes Kilometers

Short term

Short term

Short term

# Total of traffic fatalities in territory

Average speed in principal ways

Timewasteinpeakhours Timewasteinflathours

Short term

Short term

Short term

#Stationwithrealtimeinfo ∗ 100 #TotalStation #electricvehicleschargingstations #Totalregisteredelectricvehicle ∗ 100

Short term

#Publictransportocupationperlocation ∗ 100 #Totalciticenzperlocation

Short term

#PublicCleanVehicle ∗ 100 #Totalpublicvehicle

#Publictravelspercapita

Applicability period

Measure

Table 4. KPIs comparison and applicability period in Colombia (Source Authors)

A Novel Approach for Smart Mobility Roadmap in Smart Cities 177

SC-ISO-2

SC-ISO-6

SC-ISO-7

SC-ISO-8

SC-ISO-12

SC-ISO-10

SC-ISO-11

SC-ISO-13

18

19

20

21

22

23

24

25

SC-ITU-7

SC-ITU-3

SC-ISO-9

KPI SC ITU ID

17

KPI SC ISO ID SC-ITU-4

KPI SC COL ID

16

No

Medium term Medium term Medium term Medium term Medium term Medium term Long term Long term

#IntelligentTrafficlights ∗ 100 #Totaltrafficlights #Sharingtransportationusers 10.000people #PublicTransportationwithunifiedsyst ∗ 100 #Totalpublictransportation #Publicparkingwithepaymentsyst ∗ 100 #Totalpublicparking #Publicparkingwithrealtimeavailability ∗ 100 #Totalpublicparking #Publictransportationwithinternetconnectivity ∗ 100 #Totalpublictransportationlines #areamappedbyrealtimeinteractivestreetmap ∗ 100 #CityTotalarea #Autonomuosvehicles #Totalregisteredvehicle ∗ 100

Long term

Medium term

#Lengthofpublictransportnetwork 10.000people

#kilometresofroadwithautonomoussystems ∗ 100 #Totalkilometersofroad

Applicability period

Measure

Table 4. (continued)

178 T. Zona-Ortiz and G. Guzman

A Novel Approach for Smart Mobility Roadmap in Smart Cities

179

As a result, the set of KPIs allows great flexibility to structure a smart mobility roadmap, thanks to the fact that 52% of it consists of KPIs that are not considered in the MINTIC initiative, and the coincidences of the KPIs proposed by the three organizations studied are low.

4 Smart Mobility Roadmap Approach Building a Smart City project is cooperative work where a lot of stakeholders who need to be satisfied are involved. Therefore, efficient communication and information exchange inside an ecosystem is very important [18]. Smart Mobility is one of the ecosystems in a smart city. It is made up of people, entities, technological solutions, and government, among others. All these actors interact within the ecosystem generating a lot of information and data, which must be analyzed to make decisions and plan the city. Smart Mobility as the ecosystem in a Smart City should have an evolution. To plan this evolution a roadmap approach is presented, and KPIs measurement will be the tool to evaluate it. The city evolution has different stages: the city begins as an existing city, which means it has citizens and basic infrastructure; then services based on technological solutions are implemented over that basic infrastructure to collect information, so the city becomes a Digital city; then when the city makes decisions based on data from the collected information, the city becomes a Smart city; and finally, when those decisions are aligned with the Sustainable Development Goals (SDGs) and national policies to improve the city services and quality of life, the city is transformed in a Sustainable City [18]. Although a set of KPIs was established to evaluate the city’s evolution in Smart Mobility, the roadmap must consider the stages above and the gaps present in the Colombian territory. Therefore, the roadmap presented is tailored for a large city like Bogotá. Depending on the specific case, a reduction in the number of KPIs within the roadmap may be considered or an extension of the execution period may be necessary. Moreover, in some cities, a Smart City project should begin with the development of the basic infrastructure, so it is a relevant limitation for Smart City implementation. The roadmap approach is divided into three applicability periods: short, medium, and long-term. Each term includes the necessary solution implementation to collect the KPI data, data normalization and socialization, and at least one-year KPI measurement. The period to execute one applicability period in the roadmap depends on the number of KPIs considered to be implemented in that period. In this approach, 5 KPIs implementation by the term is considered. Hence short-term is defined as 5 years per 5 KPIs group for a total of 15 years; the medium-term is defined as 10 years; and the long-term is defined as 20 years. Table 4 categorizes KPIs in the defined terms, thirteen KPIs could be implemented in the short term, nine in the medium term, and three in the long term, these terms subdivisions are based on the normal evolution of infrastructure and technological solutions in the city. According to assumptions, it seems that the implementation of the set of KPIs could take 55 years. Figure 3 presents the roadmap approach for Smart Mobility in a Smart City, this deployment highly depends on the city budget, policies priorities, experience, and of course public adoption. Depending on these factors and how they are managed, the

180

T. Zona-Ortiz and G. Guzman

Data lake and big data techniques to make decision

Sustainable City

Data collection to Smart Mobility Decisions and Storage

Smart City

Evolution

Other verticals

Other verticals

Short term

Medium term

KPIs No

Smart Mobility KPIs Implementation

1 2 3

7,8,9,10,12 1,2,3,6,12 4,5,11 15 Years

Long term

KPIs No 1 2

14,16,19,20,21 15,17,18,22

KPIs No 1

23,24,25

20 Years

Digital City

City

Smart Mobility KPIs Measurement

20 Years

Implementation of set o KPIs takes 55 years

Basis Infrastructure and Citizen

Existing City

Fig. 3. Proposed roadmap approach for smart mobility in a smart city. (Source Authors)

deployment of a roadmap may have some limitations in time, budget, scope, and adoption. The roadmap is proposed to be implemented in a big city and includes the entire set of KPIs, but in smaller territories with low budgets, the priorities to improve quality of life are not related to the collection of the KPIs data. This limitation should be faced in the framework and considered in the roadmap through the assumption to plan it. Therefore, this novel approach is a guideline to plan any roadmap in Smart Mobility. To reach a digital city stage, it is necessary to offer several services that collect the needed data for building a KPI. It does not matter if those services use manual or automatized interfaces to store data in a Database, or if they deploy any infrastructure to collect data without human interaction. All KPI implementations require the collection of data from various services, regardless of the duration of the implementation period. Highlight that KPI to be implemented should consider not only the availability of data but also the probability that the KPI can be easily affected. The Smart City stage is reached when it is possible to make decisions based on a KPIs value in a vertical. That is why it is necessary to have at least two KPIs implemented in at least one vertical. The city must have storage infrastructure, and databases should have interfaces like APIs o ETLs to facilitate services and data interoperability and convergence. The final stage, Sustainable City is reached when correlations between KPIs from different verticals can be found and it is used to make decisions for city planning. As a result of these decisions, change in the KPIs values occurs. It is important to highlight that a continuous measure of the implemented KPIs in the city is mandatory at this stage.

A Novel Approach for Smart Mobility Roadmap in Smart Cities

181

5 Conclusions This novel approach presents a method to build a roadmap in Smart Mobility based on a set of 25 KPIs, based on KPIs for smart cities proposed by two international organizations and a national referent. The roadmap suggests that each KPI within the set could be implemented in a specific applicability term, considering a linear approximation for fixed applicability terms for 5 KPIs, as well as the necessary evolution from an existing city to a smart and sustainable city. Therefore, the execution period to implement the full set of KPIs in the roadmap is estimated to take around 55 years. It is important to highlight that this execution period depends heavily on the number of KPIs to be implemented, the city budget, policy priorities, and public adoption, Furthermore, the main limitations are the deployment experience and the city’s basic infrastructure state, because it is the starting point implementation. For future works, it could be considered to apply this method for other vertical roadmaps that are part of the Smart and Sustainable Cities initiative. Additionally, it is important to articulate different verticals to make decisions and plan the city. Another factor that could lead to the implementation of a roadmap is the promotion of ICT projects that enhance the city’s ability to collect data and make decisions, this strategy must be aligned with framework deployment.

References 1. Zona-Ortiz, A.T., Fajardo-Toro, C.H., Pirachicán, C.M.A.: Proposal for a General Framework for the Deployment of Smart Cities Supported in the Development of IoT in Colombia. Iberian J. Inf. Syst. Technol., no. RISTI, N.º E28, 04/2020, 894–907 (2020) 2. United Nations: Goal 11: Make cities inclusive, safe, resilient and sustainable. 25 May 2022. [Online]. Available: https://www.un.org/sustainabledevelopment/cities/ 3. Gassmann, J.B.O., Palmi´e, M.: Smart cities : introducing digital innovation to cities. https:// books.google.com/books/about/Smart_Cities.html?id=V_olwAEACAAJ, p. 341 (2019) 4. ITU-T Focus Group on Smart Sustainable Cities: Standardization roadmap for smart sustainable cities. In: International Telecommunication Union, Geneva, Switzerland (2015) 5. ISO: Sustainable cities and communities — Descriptive framework for cities and communities, ISO 37105:2019. ISO, Geneva, Switzerland (2019) 6. ISO: Smart community infrastructures — Common framework for development and operation, ISO/TR 37152:2016. ISO, Geneva, Switzerland (2016) 7. ISO: Framework for integration and operation of smart community infrastructures. Part 1, ISO 37155–1:2020. ISO, Geneva, Switzerland (2020) 8. ISO: Framework for integration and operation of smart community infrastructures. Part 2, ISO 37155–2:2021. ISO, Geneva, Switzerland (2021) 9. ISO: Report of pilot testing on the application of ISO smart community infrastructures standards, ISO/TR 37171:2020. ISO, Geneva, Switzerland (2020) 10. Wang, M., Zhou, T.: Does smart city implementation improve the subjective quality of life? Evidence from China. Technol. Soc. 72(102161), 1–13 (2023) 11. Freitas A., Brito L., Baras, J. S. K.: Smart mobility: A survey. Internet of Things for the Global Community, IoTGC 2017. doi: https://doi.org/10.1109/IOTGC.2017.800897 12. Ortiz, V., Pardo, H.: Universidad Pontificia Bolivariana. 2021. [Online]. Available: https://rep ository.upb.edu.co/bitstream/handle/20.500.11912/9609/238_1%20(1).pdf?sequence=1

182

T. Zona-Ortiz and G. Guzman

13. Burgos, S.V.: Establecer la articulación entre los indicadores internacionales y nacionales propuestos para la medición de desarrollo de Ciudades Inteligentes. Bogotá.: Universidad Santo Tomás (2022) 14. Cristina Bueti, D.C.: Collection Methodology for Key Performance Indicators for Smart Sustainable Cities. United 4 Smart Sustaintable Cities, Geneva (2017) 15. ISO: INTERNATIONAL STANDARD ISO/FDIS 37122 Sustainable cities and communities Indicators for smart cities, Geneva: ISO (2019) 16. Iván Mauricio Durán, I.T.M.L.F.M.J.C.O.A., Camaccho, M. A., Arévalo, J. D. C.: MODELO DE MEDICIÓN DE MADUREZ DE CIUDADES Y TERRITORIOS INTELIGENTES PARA COLOMBIA. MINTIC COLOMBIA, Bogotá (2021) 17. MINTIC COLOMBIA: Borrador Modelo de Madurez de Ciudades y Territorios Inteligentes (2019) [Online]. Available: www.mintic.gov.co. [Accessed December 2020] 18. AtmakurI, V.V., Lari, M.S.N., Thangaraj, A.: Design for sustainable smart cities; an impactful approach through the role of designers towards future of mankind. In: Design for TomorrowV1 Proceedings of ICoRD 2021, Mumbai, India, Springer, pp. 961–970 (2021)

Exploring the Use of Digital Twin in Smart Healthcare: A Case Study of Dengue Epidemic Control and Prevention Andres Rey Piedrahita(B) , Jenniffer Castellanos-Garzón , Julián Eduardo Betancur, Marco Tulio Canizales, Juan Sebastián Henao-Agudelo, Luis Alberto Rivera Martinez , and Sebastian Lopez-Mejia Unidad Central del Valle del Cauca, 08544 Tuluá, NJ, Colombia [email protected]

Abstract. Dengue is a disease that spreads through mosquito bites and is contagious. It mainly occurs in tropical and subtropical regions, putting about a third of the world’s population at risk of getting the disease. Due to the high number of deaths caused by Dengue outbreaks, governments have developed policies to combat and prevent the disease. However, the available data on reported cases suggest that a deeper understanding of the complex phenomenon is necessary to improve control and prevention in endemic regions. Using dengue as a use case, this article is a progress report on the use of digital twins in smart healthcare. The use of a model of system dynamics in the context of digital twins is offered as a novelty. This technology allows for the assessment of the long-term effects and results of different policies, as well as the identification of the underlying causes of the disease’s behavior. The following results of the initial work in progress are presented: A system dynamics model of the complex system involved in the control and prevention of dengue, which will guide the data collection and form the basis for the subsequent construction of the digital twin; a high-level outline of the architecture that will be implemented to deploy this resource. It is hoped that the subsequent implementation of the Digital Twin will enable intelligent experimentation with public policy against infectious diseases such as dengue. Keywords: Smart health management · Digital twin · Control and prevention · Dengue epidemic · Public politics

1 Introduction Dengue fever is an infectious disease caused by infection with one of four types of dengue virus (DENV 1–4). Female Aedes aegypti mosquitoes are the primary vectors of the virus and transmit it to humans. The disease is most prevalent in tropical and subtropical regions, where approximately one-third of the world’s population is at risk of contracting the disease [13]. Several reports suggest that the incidence of infection has dramatically increased by more than 30 times in the last five decades [24].

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 183–193, 2023. https://doi.org/10.1007/978-3-031-36957-5_16

184

A. R. Piedrahita et al.

The World Health Organization (WHO) estimates that approximately 390 million people have been affected by the disease, with more than 96 million new cases occurring each year. However, the actual impact of the disease in developing countries is not clear [6]; see Fig. 1.

Fig. 1. Annual cases of dengue in the Americas, 1980–2019 (Source PLISA: Health Information Platform for the Americas)

Dengue virus infection can cause a range of pathological conditions, including dengue without warning, dengue with warning and severe dengue, which can lead to severe outcomes such as dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS). These severe symptoms can be fatal and can also lead to hepatitis, neurological disorders and myocarditis [6, 13, 24]. If not properly treated, severe dengue can have a mortality rate of up to 20%; however, with proper care and management of rehydration, the mortality rate can be reduced to 1% [12]. It is therefore essential that doctors working in endemic areas have expertise in the treatment of this arbovirus, and that communities engage in prevention and knowledge-sharing activities. To address this situation, governments are formulating policies to combat the disease, such as influencing the reproduction of the mosquito to reduce the incidence of dengue. Campaigns are also carried out to socialise prevention measures in the community. However, the different data reported on the subject show that it is necessary to advance in the understanding of the complex problem that is configured in relation to the disease. On the other hand, in recent years, the development of the Internet of Things (IoT) and the proliferation of sensor and connected actuator technologies have changed the way data is shared between different sources, leading to an increase in the creation of large data sets. Thanks to advances in big data analytics, cloud computing, modelling and simulation (M&S), and the incorporation of artificial intelligence (AI), it is now possible to store and analyse IoT data, creating opportunities for the advancement of smart health management. These systems can take advantage of Digital Twin (DT). However, accurate DT development requires a comprehensive understanding of each

Exploring the Use of Digital Twin in Smart Healthcare

185

part within a system, the relationship between these parts, and the analytical power to evaluate the effect of variables introduced into the system [2]. Using dengue as a use case, this article is a progress report on the use of digital twins in smart healthcare. The first results of the work in progress are presented: a system dynamics model of the complex system involved in the control and prevention of dengue, which will guide the data collection and form the basis for the subsequent construction of the digital twin; and a high-level schematic of the architecture that will be implemented to implement this resource.

2 Background 2.1 Digital Twin The original Digital Twin concept is defined as a digital representation of a physical system that exists independently and is connected to the physical system it represents [2]. Ideally, this digital model should contain all the information about the physical asset that could be gathered from its thorough inspection in the real world, as stated by [11]. Initially, the digital twin concept referred to the creation of a virtual representation of a product. However, as technology has advanced, it has become possible to apply the same concept to processes and systems such as manufacturing, power generation, healthcare systems and cities. By creating a virtual replica or ‘twin’ of these processes or systems, it is possible to gain the same benefits as with product twins [26]. Digital Twin refer to virtual versions of physical objects, and the terms digital model, digital shadow and digital twin are often used interchangeably. However, the definitions of these terms can vary depending on the level of data integration between the physical and digital versions. Some digital models are manually created and not linked to existing physical objects, while others are fully integrated and allow for real-time data exchange [14]. 2.2 The Modelling and Simulation of Complex Systems According to Yoon et al. [27], a complex system is made up of interconnected components or parts that interact with each other. Such systems exhibit emergent properties that cannot be observed by looking at the individual parts, but arise from the structure created by the relationships between them. When perturbations occur, the self-organisation of the system’s structure can lead to the emergence of unpredictable new properties, as noted by Barelli and Laverini [5] and Yoon et al. [27]. Examples of complex systems include cities, transport or communication systems, living organisms and the human brain, infrastructures such as electrical networks or electronic systems, the Earth’s climate, social and economic systems, the health system, and more [3, 7, 10, 22, 23]. When dealing with complex systems, science uses modelling and simulation techniques [8]. These tools allow the study of real-world phenomena or objects that are difficult to observe or study in a controlled laboratory environment. In addition, simulation allows for the solution of complex mathematical problems that would be impractical

186

A. R. Piedrahita et al.

to solve manually due to the large number of calculations required [17]. The literature suggests that working with models enables the study of how the real world works by dealing with complexity and uncertainty [18]. Typically, the process involves first gaining a deeper understanding of the dynamics of real systems and the challenges they pose, and then improving the modelling and control of these systems and problems. 2.3 Relate Work This originated from mirrored systems or simulated environments created by NASA in the 1970s to monitor unreachable physical spaces (for example, the spacecraft of the Apollo 13 mission) [4]. Currently, Digital Twin are used in various industries such as manufacturing, construction, automotive, and aerospace. These applications include tasks such as product design, service management, product lifetime prediction and realtime equipment monitoring [2]. In addition, there is an emerging trend towards the use of DT in healthcare and the management of healthcare systems [9]. In terms of precision medicine, DT have significant potential as they can be used to simulate personalised therapies and visualise possible outcomes and disease progression for individual patients; a review of the literature on this topic is presented in Ref. [2]. Furthermore, DT can be used at the system level, offering a wide range of applications, from ensuring safety [1] and managing information to promoting health and well-being [15] and controlling operations [19]. Nevertheless, the use of DT in the management of healthcare systems is an evolving area with enormous potential [9]. It is concluded that the use of DT in health care is a recent development compared to other fields, according to the available literature.

3 Methodology The application exploration of DT in Smart Health Management is carried out for the above case: Dengue epidemic control and prevention in an endemic region. The development of a prototype system. The materials and methods used are described below. 3.1 Materials and Methods Bearing in mind that an accurate DT development requires a broad understanding of each part within a system, the relationship between these parts, and the analytical power to evaluate the effect of variables introduced into the system [2]. It is determined to build a white box model (represented by a causal loop diagram, CLD) for the DT modelling the case of dengue control and prevention in endemic regions. The developed model provides a representation of the structure of the real system involved in the control and prevention of dengue in endemic regions of Colombia. A structural validation of this model is carried out by directly comparing its structure with the existing knowledge about the structure of the real system: the aim is to determine if the model, at the structural level, represents a conformity with physical realities and known basic laws. For this test, the existing theoretical knowledge in the scientific literature

Exploring the Use of Digital Twin in Smart Healthcare

187

about the system is used, as well as quantitative or qualitative information obtained directly from the case study to be modelled [25]. System Dynamics (SD), a technique for modelling and simulating continuous events, is used to construct the model for the DT. It has the potential to effectively represent the internal structure of complex systems, where decisions and human actions over time contribute to observable behaviours. The literature suggests that SD is well suited to analysing complex systems involving multiple stakeholders. Previous studies have used this technique to investigate issues related to open government, such as open government data [16, 20, 21]. Initial data is gathered primarily through a thorough review of the scientific and technical literature on the subject. This information is supplemented by observation and some interviews with experts in the field. To implement the DT in the Smart Health Management technology platform, the PySD1 and sdCloud2 Project tools will be used.

4 Results 4.1 System Dynamics Model for Preventing Dengue The activities carried out to date have enabled the construction of a causal impact CLD, shown in Fig. 2. The CLD provides an understanding of the relevant variables, relationships and cycles in the dynamics of dengue control and transmission in endemic regions.

Fig. 2. CLD for control and transmission of Dengue in endemic regions

In Fig. 2 several balance cycles of the balance (B), which are part of the system structure, are shown. Some of these are described below. 1 https://pysd.readthedocs.io/en/master/index.html. 2 https://sdcloud.io/.

188

A. R. Piedrahita et al.

Cycle B1 (promotion of dengue prevention): In this cycle, government interest in dengue control increases as dengue cases increase. As government interest increases, so do prevention campaigns. The latter also increases community interest in preventing the disease. As the community’s interest in preventing the spread of the disease increases, so do prevention campaigns against dengue, which means that the number of reported cases decreases; in this cycle, the government’s interest in dengue control decreases as other diseases that also require the government’s attention increase. Cycle B2 (interest to dengue prevention): In this cycle, community interest in dengue prevention grows as dengue prevention campaigns increase. This increases prevention and reduces mosquito populations. Cycle B3 (Dengue prevention): In this cycle, the mosquito population decreases as dengue prevention increases. As the mosquito population decreases, the number of cases also decreases, as the interactions between mosquitoes and the human population are reduced (epidemic model). The latter has an impact on knowledge about dengue transmission and control, which increases as the population experiences the disease and as prevention campaigns increase. Cycle B4 (Dengue control): In this cycle, government interest in dengue control increases as the number of dengue cases increases. As the government’s interest increases, so do the control campaigns (fumigation, etc.), which reduce the mosquito population and thus the number of cases by reducing the interactions between mosquitoes and the human population (epidemic model). Figure 2 also shows the system actors in relation to the model variables. These are: the government, the local authorities, the community, the mosquito, the environment. The CLD captures the actions of these actors in relation to the problem of the dengue epidemic. 4.2 Smart Health Platform Architecture An important aspect of effectively modelling a complex system using SD is the ability to capture data present in the system. In this sense, technologies such as the Internet of Things (IoT), cloud computing and machine learning become key resources that allow the development of an accurate DT. To realise the DT in Smart Health Management, an IoT network consisting of environmental sensors has been deployed throughout the territory of a university.3 These environmental variables are captured in real time and can be used as inputs for the SD model to estimate risk levels. The data collected by the IoT network is transmitted over the Internet to cloud services where it is automatically managed, processed and analysed. Using cloud computing, the DT can run estimates of risk levels from machine learning models that take as inputs the variables captured by the IoT network. The development of a DT that models the behaviour of dengue in an endemic area requires that the system that captures the environmental variables is able to scale with the 3 The Central Unit of Valle del Cauca (UCEVA), located in the city of Tulua Valle del Cauca in

Colombia.

Exploring the Use of Digital Twin in Smart Healthcare

189

expansion of the area. This is particularly important as the population in the area tends to increase over time, taking up more land. To achieve this, our DT was designed using IoT networking standards used in smart city use cases, such as the MQTT machineto-machine networking protocol, which enables a public/subscribe messaging protocol between the IoT network sensor nodes and the cloud services, simplifying message routing and device management. In addition, the computing architecture used by the cloud provider makes use of service containerisation using the Docker engine. In this way, services can be easily added and updated in the architecture, allowing the DT to evolve by incorporating services according to the needs of the system. 4.3 Implementation Progress Ongoing work focuses on: (1) development of the dengue epidemic control and prevention platform and (2) implementation of the computer simulation model for the complex system involved. 4.3.1 System Dynamics Model for Preventing Dengue The computer simulation model is in the formulation and implementation phase. Three modules will be developed: (1) a module for the population of mosquitoes and humans; (2) a simple epidemic module SIR (Susceptible, Infected, Re-covered); and (3) a prevention and health awareness module where some relevant soft variables will be included. Using the SD and based on the CLD, progress is made in constructing the stock and flow diagram (SFD). Stock and flow diagrams, also known as Level and Rate diagrams, provide a more detailed representation of a system’s structure compared to causal loop diagrams. These diagrams illustrate the fundamental role of stocks (or levels) in driving the behavior of a system, while flows (or rates) are responsible for changing these stocks. Figure 3 shows the SFD for the human population and Fig. 4 shows the behaviour over time. 4.3.2 Smart Health Platform Architecture Diagram The overall architecture of the platform on which the DT will be deployed can be observed in Fig. 5, where the components can be grouped into three categories, such as: (1) monitoring devices deployed on site, (2–4) computing services deployed on the AWS cloud, and (5–6) visualisation clients used by community stakeholders. The IoT network is made up of devices that monitor temperature and humidity at adjustable time intervals. These devices are connected to the Internet via a Wi-Fi network deployed along the territory and transmit data using the MQTT protocol. The data transmitted corresponds to the environmental values measured at the location where the devices are deployed, as well as information on the unique identification of the device, its name, coordinates and status. All data from the devices is published to an HTTP endpoint corresponding to a server that manages the IoT devices and messages. This server is an instance of the AWS IoT Core service, which simplifies the management, security and scaling of the network. Then, the messages received at the AWS IoT Core

190

A. R. Piedrahita et al.

Fig. 3. SFD of the human population variable in the simulation model. The population growth rate determines how the population changes over time. The poverty index makes it possible to calculate the population with satisfied and (un)satisfied basic needs.

Fig. 4. Behaviour of the human population variable generated in Powersim4 simulation software. The growth of the total population (grey line), the proportion with (un)satisfied basic needs (blue line) and the proportion with satisfied basic needs (orange line) is shown.

service are transferred to two instances of the AWS Lambda service: the first (3.a), where a Python code transforms the data schema and sends it to a database server (4) that stores the time series of these variables; and the second (3.b), where the data is transformed and could be routed to an instance of the PySD library that runs the SD model of the complex system to perform the estimation of the risk levels over the area. The estimated levels are then stored in the time series database (4). The information captured and processed by the architecture enables the estimation of the SD model of the territory and becomes an enabling resource for the implementation of real-time data-driven approaches to health management.

4 https://powersim.com/.

Exploring the Use of Digital Twin in Smart Healthcare

191

Fig. 5. Platform for the Digital Twin in the Smart Healthcare Platform

5 Discussion and future work Using dengue as a use case, this article is a progress report on the use of digital twins in smart healthcare. The first results of the work in progress are presented: a system dynamics model of the complex system involved in the control and prevention of dengue, which will guide the data collection and form the basis for the subsequent construction of the digital twin; and a high-level schematic of the architecture that will be implemented to implement this resource. Regarding the model developed, it is worth mentioning that it is a white box model (represented by a causal loop diagram) of the complex system configured around the spread of this disease. This model was created because the accurate development of DT requires a broad understanding of each part within a system, the relationship between these parts, and the analytical power to evaluate the effect of variables introduced into the system [2]. The model developed with the System Dynamics Methodology (based on the case study of dengue in Valle del Cauca, Colombia) provides a representation of the structure of the real system involved in the control and prevention of dengue in endemic regions. From the CLD of the system, an SFD will be built to implement the digital twin. This will later make it possible to experiment with different policies and strategies formulated around this disease, to know their effects and long-term results, in addition to identifying the different causes of the behaviour presented; the CLD is also an artefact that facilitates communication with stakeholders. This study also reports on some tools, such as PySD and sdCloud Project, that make it possible to work with system dynamics models outside the typical simulation environments, such as Stella/I Think, Powersim and Vensim, and to move them to other platforms; this is useful for a group of practitioners of this discipline who are not very used to this type of deployment.

192

A. R. Piedrahita et al.

The Digital Twin will be deployed on a Smart Health Management technology platform currently under development. This will enable the collection, storage and processing of data. The knowledge gained from the data will be used to adapt and improve the system model. This will also allow us to test some of the CLD model’s assumptions about the causes of the disease, such as population growth, knowledge and interest in prevention, and changes in environmental conditions that favour the mosquito’s reproduction and spread. In conclusion, it can be argued that simulation models of complex systems, together with other technologies, can be useful in building DT as a key enabler for digital transformation, and in the area of Smart Health Management, the work is just beginning. Acknowledgment. The results presented are part of the project "Multidisciplinary and Integrated Dengue Study - EMID: Evaluation of epidemiological, clinical, immunobiological, genetic, sociodemographic, community and health factors associated with severe Dengue in an endemic region", funded by Call 890 of Min-Ciencias of 2020.

References 1. Alrashed, S., Min-Allah, N., Ali, I., Mehmood, R.: COVID-19 outbreak and the role of digital twin. Multimed. Tools Appl. 81(19), 1–15 (2021). https://doi.org/10.1007/s11042-021-116 64-8 2. Armeni, P., Polat, I., De Rossi, L.M., Diaferia, L., Meregalli, S., Gatti, A.: Digital twin in healthcare: is it the beginning of a new era of evidence-based medicine? A critical review. J. Pers. Med. 12(8), 1255 (2022) 3. Arroyo, M., Hassan, S.: Simulación de procesos sociales basada en agentes software. Empiria Revista de Metodología de las Ciencias Sociales 14, 139–161 (2007) 4. Barricelli, B., Casiraghi, E., Fogli, D.: A survey on digital twin: definitions, characteristics, applications, and design implications. IEEE Access 7, 167653–167671 (2019) 5. Barelli, E., Levrini, O.: Computational simulations at the interface of physics and society: a teaching-learning module for high-school students. Nuovo Cimento 45(6), 1–4 (2022) 6. Bhatt, S., et al.: The global distribution and burden of dengue. Nature 496(7446), 504–507 (2013) 7. Casaboza, J., Cárdenas, D.: Analysis and modeling of dynamic behavior of the COVID-19 outbreak: study case of Panama. IEEE Lat. Am. Trans. 19(6), 893–900 (2021) 8. Chattopadhway, S., Roy, T., Sengupta, S., y Berger-Vachon, C.: Modelling and Simulation in Science. Technology and Engineering Mathematics Conference Proceedings MS, p. 386 (2017) 9. Elkefi, S., Asan, O.: Digital twin for managing health care systems: rapid literature review. J. Med. Internet Res. 24(8), e37641 (2022) 10. Fenner, G., Lima, A., De Souza, J., Moura, J., Bezerra, T.: Supporting infrastructure as a service capacity management through business scenarios simulation. IEEE Lat. Am. Trans. 18(03), 473–486 (2020) 11. Grieves, M., Vickers, J.: Digital Twin: Mitigating Unpredictable, Undesirable Emergent Behavior in Complex Systems. In: Kahlen, F.-J., Flumerfelt, S., Alves, A. (eds.) Transdisciplinary Perspectives on Complex Systems, pp. 85–113. Springer, Cham (2017). https://doi. org/10.1007/978-3-319-38756-7_4

Exploring the Use of Digital Twin in Smart Healthcare

193

12. Guzman, M., Gubler, D., Izquierdo, A., Martinez, E., Halstead, S.: Dengue infection. Nat. Rev. Dis. Prim. 2(1), 1–25 (2016) 13. Khetarpal, N. Khanna, I.: Dengue fever: causes, complications, and vaccine strategies. J. Immunol. Res. (2016) 14. Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W.: Digital Twin in manufacturing: a categorical literature review and classification. Ifac-PapersOnline 51(11), 1016–1022 (2018) 15. Liu, Y., Zhang, L., Yang, Y., Zhou, L., Ren, L., Wang, F. et al.: A novel cloud-based framework for the elderly healthcare services using digital twin. IEEE Access 49088–49101 (2019) 16. Luna-Reyes, L., and Ramon Gil-Garcia, J.: Using institutional theory and dynamic simulation to understand complex e-Government phenomena. Gov. Inf. Q. (28), 329–345 (2011) 17. Maldonado, L.: El modelamiento matemático en la formación del ingeniero. Ediciones Universidad Central, Bogotá (2013) 18. Medina, J.: Explicación, mecanismo y simulación: otra manera de hacer sociología. Empiria: Revista de Metodología de Ciencias Sociales (28), 35–58 (2014) 19. Mylrea, M., Fracchia, C., Grimes, H., Austad, W., Shannon, G., Reid, B. et al.: BioSecure digital twin: manufacturing innovation and cybersecurity resilience. Eng. Artif. Intell. Syst. 53–72 (2021) 20. Najafabadi, M., and Luna-Reyes, L.: Open Government Data Ecosystems: A Closed-Loop Perspective. In: Proceedings of the 50th Hawaii International Conference on System Science 2711–2720 (2017) 21. Najafabadi, M.: Modeling an Open Data Ecosystem: The Case of Food Service Establishments Inspection in New York State. State University of New York at Albany (2020) 22. Ncube, C., and Lim, S.: On systems of systems engineering: a requirements engineering perspective and research agenda. In: IEEE 26th International Requirements Engineering Conference, pp. 112–123 (2018) 23. Parhizkar, T., and Mosleh, A.: Guided probabilistic simulation of complex systems toward rare and extreme events. In: Annual Reliability and Maintainability Symposium, pp. 1–7 (2022) 24. Rather, I., et al.: Prevention and Control Strategies to Counter Dengue Virus Infection. Front Cell Infect. Microbiol. (2017) 25. Sterman, J.: Business dynamics: systems thinking and modeling for a complex world. McGraw-Hill, Boston (2000) 26. Tao, F., Cheng, J., Qi, Q., Zhang, M., Zhang, H., Sui, F.: Digital twin-driven product design, manufacturing and service with big data. Int. J. Adv. Manuf. Technol. 94(9–12), 3563–3576 (2017). https://doi.org/10.1007/s00170-017-0233-1 27. Yoon, S., et al.: Designing computer-supported complex systems curricula for the Next Generation Science Standards in high school science classrooms. Systems 4(38), 2–18 (2016)

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain Mohamed Ayman Abu El Magd1(B) , Nada Sharaf2 , and Maggie Mashaly1 1

The German University in Cairo, Cairo, Egypt {mohamed.abu-elmagd,maggie.ezzat}@guc.edu.eg 2 The German International University, Cairo, Egypt [email protected]

Abstract. In recent years, cryptocurrency has made a significant impact on the financial industry, primarily due to its secure and immutable nature enabled by blockchain technology. However, the use of blockchain is not limited to finance and can be applied to various fields, including supply chain management. This article presents an introduction to the topic of blockchain technology and its potential application in supply-chain management. Research conducted in this project employs Hyperledger Fabric to create a fully functional blockchain network capable of managing a large supply chain. The proposed architecture has two channels, three peers, and one orderer organization, each representing different stages of the supply chain. The system was tested by installing various chain codes to evaluate its effectiveness in tracing transactions and recording the product’s life cycle. Nevertheless, the system has some limitations and requires further development to be ready for full-scale deployment. This prototype serves as a proof-of-concept for a blockchainbased supply chain management system. Keywords: Blockchain · Hyperledger Fabric · Ownership transfer · Peer-to-peer network · Supply chain management · Tracing transactions

1

Introduction

With the COVID-19 pandemic forcing people to work and shop from home, cryptocurrencies like Bitcoin gained more attention, and the blockchain system responsible for securing their value became a topic of interest [11]. This paper discusses the potential of blockchain technology to extend beyond financial applications and into supply-chain management. To this end, a project is proposed to create a small network secured by blockchain to trace products efficiently and cost-effectively using open-source blockchain networks. Supported by The German University in Cairo c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 194–206, 2023. https://doi.org/10.1007/978-3-031-36957-5_17

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

195

Since retailers are being more severe about enforcing delivery dates, which is why prompt product delivery is so important to the supply chain, wholesalers have started investing in technology and training. It may be difficult to create confidence amongst supply chain partners, especially for organizations with diverse locations in the world, when there are delays in product delivery. Although ERP systems have attempted to solve the issue, third-party accountants are frequently controversial and unreliable, and information movement between various organizations in the supply chain is difficult. Because of its transparency and data immutability features, blockchain technology is seen as a suitable option for supply chain transactions [12]. This research paper further explores the potential applications of blockchain in supply chain management and evaluates the feasibility of the proposed project, with the goal of contributing to the literature on the subject and providing insights for future research by designing an architecture that could be used to fill the gap of involving several institutions in trustless network and connecting them on one single platform. The rest of the paper is organized as follows: Sect. 2 Background shows more information about blockchain and its relation with supply chain management, Sect. 3 Methods which shows the steps of the technical work done to achieve the purposed system architecture, Sect. 4 Results shows the resulted information on the backend side of the network and Finally, Sect. 5 Conclusion that summarizes the paper and discusses the suggested future work for the project.

2 2.1

Background Blockchain

Blockchain is a secure and decentralized method of recording transactions that uses immutable cryptographic signatures to prevent alteration or fraud. Each transaction is added to a ledger that is copied and distributed throughout the network, making it nearly impossible to hack or tamper with [10]. Here are the steps involved in a blockchain system: 1. A user initiates a transaction. 2. The transaction is broadcasted to all nodes on the network. 3. Network validates transactions to prevent fraud with the principles of blockchain immutability. 4. Block created with validated transactions. 5. Validators are rewarded through transactions’ tips (like Ethereum) or mining and tips (like Bitcoin). 2.2

Types of Blockchain

A public blockchain is a decentralized technology that distributes data over a peer-to-peer network, eliminating centralization’s drawbacks. It requires consensus algorithms to authenticate data and can be accessed by anyone, offering transparency, independence, and safety. It is useful for creating permanent

196

M. A. Abu El Magd et al.

records and is ideal for organizations that value transparency and trust, but not for private firms due to its lack of scalability [2]. Private blockchains are a restricted form of blockchain that is controlled by a single company, offering privacy and scalability benefits. Although centralized, private blockchains allow chosen participants transparency, trust, and security. However, the centralized decision-making process is a drawback, which can undermine the consensus process and compromise security. Private blockchains are ideal for internal purposes within organizations and can be used for supply chain management, asset ownership, and internal voting. Examples of private blockchain networks include Multichain and Hyperledger Fabric [3]. A hybrid blockchain combines the advantages of both private and public blockchains, allowing for both private and public permission-based systems. The users can manage access to the data stored on the blockchain. The hashing power of public blockchains increases security and transparency, and it’s resistant to 51% of attacks. Hybrid blockchains have several use cases such as IoT, finance and trade, banking, and enterprise services. Transaction costs are low, and businesses don’t have to worry about data exposure, making it an ideal solution for many industries [4]. 2.3

Different Types of Consensus Algorithms

To update the ledger, the network needs to use a consensus algorithm where all nodes agree on the current state, ensuring no double-spending occurs. Consensus algorithms are used to agree on commands/logs in distributed systems, with three major types used in blockchain networks. The Proof-of-Work (PoW) consensus system rewards miners who solve mathematical challenges to validate new network transactions. Each mining client collects incoming transactions in a memory pool and verifies and groups them into blocks. Miners have a financial incentive to be the first to produce a new block and add it to the decentralized copy of the blockchain they keep. This is achieved by solving a hard puzzle that requires a lot of computational power. The weight of each node is determined by its computational power, and the probability of being picked to post the block into the chain increases with computational power. Bitcoin’s PoW algorithm requires miners to attach a random number (Nonce) to the transaction data in their blocks as they compile them and locate at least X leading zeros in the resultant hash through trial and error. The complexity of the puzzle is dynamically changed, ensuring that the average block interval remains relatively consistent at around 10 min [1]. The current stage of blockchain development still relies heavily on the proofof-work mechanism used in Bitcoin, which consumes a lot of energy and has scalability issues. New consensus protocols, such as proof-of-stake, are being researched to increase efficiency without depending on private blockchains or expensive node vetting. Proof-of-stake is based on owning a stake in the network and nodes with more money invested to have a better chance of winning the block. This eliminates the need for mining hardware and energy. However, proof-

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

197

of-stake systems can be complicated with voting levels and may not always be more efficient than Bitcoin [8]. 2.4

Supply-Chain Relation with Blockchain

Walmart has collaborated with IBM to develop a food tracking system using blockchain technology, specifically Hyperledger Fabric. The system digitizes the food supply chain process, making it easier to trace the origins of food and identify issues with contaminated batches. Walmart and IBM worked on the system’s architecture, configuration, and interaction with enterprise systems, while barcode and labeling standards were determined by another authority. Blockchain technology was tested through two Proof-of-Concept initiatives in the United States and China, showing positive results in both cases. The system has reduced the time it takes to trace mango origins from seven days to 2.2 s. Walmart believes the technology is ready for widespread adoption, and suppliers only need to know how to upload data to benefit from the system. The system can now track the origins of over 25 goods from five different vendors [6].

3

Methods

The section covers the phases and workflow of a project related to Hyperledger Fabric. It includes sections on system design, supplementary tools, Orderers and their settings, Peers and their settings, and channels formed and their settings. 3.1

System Design

The purposed system design is as follows: the system will contain three main organizational types: the Producer, the Distributor, and the Retailer. The Producer will provide the distributor with the products that’s why they need to have a channel between them so that they can both view the data that they both share. In addition, the distributor will provide the retailer with the products that’s why they need an extra channel between them too. This design purposes some sort of a hierarchy in the system so that not all the involved entities will have access to all the data so that each distributor will have access only to its channel with the producer and will not be able to access data of the other distributor and same goes for each retailer with the distributor. As for the orderer, it will be

198

M. A. Abu El Magd et al.

Fig. 1. Configtx.YML file

an entity that will be connected to both channels for now for its importance in maintaining the network. The input will come from the user or the client (or from IoT sensors) as transactions that need to be inserted into the system. They will pass by the orderer nodes first and based on the transaction type, they will be forwarded to be published on the corresponding channel. This design can be illustrated in Fig. 1. 3.2

Hyperledger Fabric Binaries (Tools)

Cryptogen is a tool used for generating cryptographic content for testing purposes only, not for production use. The tool generates certificate and Keystone decryption crypto material using a YAML configuration file that specifies a list of orderers and peers, along with their respective features such as Specs and User. The administrator uses certificates and keystones found in the admin directory to start or stop the peer binary, install and restart the chain, and make configuration changes. This setup is only for testing and will not be used in a production release. Configtxgen is a tool for managing network/channel configuration and generates three types of artifacts: Genesis Block, Channel Tx, and Anchor Peer Tx. The tool uses configtx.YAML file to create these artifacts, which can be generated using output commands. Inspection commands are used to review the information contained in the configuration artifact, which is in binary format, by converting it to JSON format. The configtx.YML file has 6 main sections as shown in Fig. 2:

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

199

Fig. 2. System Architecture

1. Organizations: It includes a list of member organizations, orderers, and peers with their different aspects and combinations. Each organization is specified with a name, ID, MSPDirectory, and policies. The importance of defining the anchor peer is highlighted, particularly for non-orderer organizations with many peers. The project includes one orderer organization and three non-orderer organizations, each with only one anchor peer. 2. Orderer: The orderer setup settings, including the orderer type (Solo in this case), the addresses of the orderer nodes, the batch timeout (2 s), and batch size settings such as maximum message count (10), absolute maximum block size (50 MB), and preferred maximum block size (512 KB). The difference between the absolute and maximum sizes is that the block will only load transactions up to the preferred maximum size, but will load a single larger transaction if it falls between the preferred and absolute maximum sizes.

200

M. A. Abu El Magd et al.

3. Application: The list of organizations involved in transactions and the Access Control List for managing resource access by linking a Policy with a resource. The ACL provides the policies necessary for each chain code function to match the policies specified for the channel organizations. 4. Channel: It defines the set of policies for each channel 5. Capabilities: It’s responsible for binary version management. It states which version each network artifact is using. It is set to version 2 for all the artifacts involved in the network. 6. Profile: Each profile is named and needed for generating specific configuration components. The attribute values under each profile depending on the component and reference elements from other sections of the file. The project has three subsections in this part, one for each channel and the genesis block, containing channel and organization names and properties for ordering service nodes and consortium members. The creation of the genesis block requires configurations for the orderer, orderer organizations, and consortiums. The channel transaction requires configurations for the application, consortiums, and channel ID. The transaction file is submitted by the admin of the relevant organization and is converted to JSON format for readability. 3.3

Orderers

Hyperledger Fabric differs from other distributed and permissionless blockchain networks like Ethereum and Bitcoin because it uses deterministic consensus methods and has an orderer node that handles transaction ordering, creating an ordering service with other orderer nodes. This approach ensures the finality and accuracy of each block confirmed by a peer and avoids ledger forks. Separating the endorsement of chain code execution from ordering provides Fabric with advantages in efficiency and scalability [7]. The YAML file for the orderer node has 6 sections: General, FileLedger, Consensus, Debug, Operations, and Metrics. The General section configures orderer initialization and the location of the genesis block file. FileLedger specifies the location of the ledger data, Consensus specifies properties for Orderer type “etcdraft”, Debug contains debug information control, Operations is for monitoring and alerts, and Metrics provides metrics information for third-party tools. Each section is critical in configuring subsequent subsections. The File Ledger section of the YAML file configures the storage location of the ledger data, with two characteristics to set: location and prefix. If the location is specified, the prefix is ignored. The project provides a location in the orderer directory. The Crypto Service Provider (CSP) can be implemented through software libraries, hardware security modules, or smartcards. Hardware CSPs offer better performance and security, with a standard known as PKCS#11. The Orderer CSP can be configured in the YAML file’s General section, with options for software or hardware-based CSPs. The Orderer needs access to the MSP crypto

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

201

material to verify incoming messages and to sign communication with the private key. The MSP configuration is done in the General section with LocalMSPID and LocalMSPDir properties. In Hyperledger Fabric, assets are created using gRPC and the same notion as Protocol buffers. gRPC is a remote call technique built on HTTPS 2.0 that allows a client to call methods on a server on a separate computer as if it were a local object [5]. It uses proto buffers, which are language-neutral and platform-neutral and supports 10 programming languages. The orderer must be configured to listen to a certain IP address and port number to receive requests from peers, and parameters such as KeepAlive must be adjusted to avoid overuse of resources. 3.4

Peers

The peer is a node in the Hyperledger Fabric Network that manages the peer and network configuration through the core.yml file with its different sections. Each peer in a network requires its own core.yaml file with specific configurations. The file is divided into 6 sections which cover general properties, networking, MSP, storage path, state database, chain code, third-party integration, endpoint configurations, metrics, and usage of docker containers within virtual machines. A peer node receives CLI configurations and application connections and writes data to the file system. The core.yaml file assigns an identity to the peer and configures it to listen to multiple addresses and nodes. The MSP configurations in the file include localMspId and mspConfigPath, which must match the MSP ID in the genesis block and the ID mentioned in the channel genesis block, and correspond to the crypto material generated by the cryptogen tool. The gossip data dissemination protocol is a computer-to-computer technique based on social networks and epidemic spread. The protocol aims to synchronize all nodes without a central data repository, with hub nodes communicating and receiving data from each other. Nodes broadcast data to nearby nodes until all nodes in the network have received it [9] (see Fig. 3).

Fig. 3. Gossip Protocol

202

M. A. Abu El Magd et al.

The gossip protocol performs three functions: peer discovery and membership, discovering missing data, and helping new peers catch up. Peers must be in the same channel for the protocol to work. The leader peer receives data from the orderer organization and relays it to other peers. The protocol is configured in the peer section with bootstrap peers, heartbeat intervals, and leader election settings. The useLeaderElection and orgleader settings are mutually exclusive, and it’s recommended to use dynamic leadership election for larger networks.

4

Channels

Hyperledger Fabric channels are private subnets for confidential transactions between specific network users. Each channel comprises members, anchor peers, a shared ledger, chain code applications, and an ordering service node. The channel-based separation of peers and ledger data allows users to have private transactions on the same blockchain network as competing companies and prohibited members. The configuration for each channel is done in the configtx.yml file, and a genesis block is created for each channel. Peers can join a channel by fetching the genesis block and must be authenticated and allowed to transact on that channel. If a peer fetches a genesis block for the wrong channel, the proposal will be rejected, and the peer will not join the channel. These steps are shown in Fig. 4.

Fig. 4. Initiating and Joining a Channel

5 5.1

Results Generating the Crypto Material and Lauching Orderers

First, we start by the creation of crypto files for an OrdererOrg and three peer organizations, each with one peer node. The crypto files include a Certificate

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

203

Authority folder, an MSP folder, an Orderer folder, a Transport Layer Security Certificate Authority folder, and a Users folder. The peer organizations’ crypto folders have the same information as the orderer crypto folders, but with different organization names and are referenced in the core.yaml file of the peer. The orderer node uses the configtx.yaml file to create the network artifacts. Firstly, the orderer generates the genesis block for the network and creates a human-readable JASON file for it. Then, the orderer creates an initial transaction for each channel in the configtx.yaml file, which contains all the configurations required for the channel. The resulting files are saved in the orderer folder with the channel name, and they can be converted into human-readable JASON files. In the project, two channels result in two initial transaction files, one for Prod Dist-Channel and another for Dist Ret-Channel. The orderer node launches its service and starts serving requests from other ports after reading and compiling the configurations specified in the orderer.yaml file, which is generated after creating all the necessary files as shown in Fig. 5.

Fig. 5. Orderer Launching

5.2

Launching Peers

The peer nodes can be launched by the running orderer node using the configurations specified in the core.yaml file. Once launched, the peer node will start accepting requests from client servers for chain code invocation, installation, and packaging, after joining the channels as specified in the configtx.yaml file. The peer node will also configure the leader peer and the anchor peer based on configurations and encrypted files, and the leader peer may be elected dynamically. Only one peer node per organization can be an anchor peer, and the launching process is detailed in Fig. 6a and b.

204

M. A. Abu El Magd et al.

(a)

(b)

Fig. 6. Launching Peers

5.3

Joining Channels

The peer node must join the channels specified in the configtx.yaml file, starting with the creation of the genesis block for the channel if it does not already exist. The genesis block is an encrypted file, and the first peer to join the channel creates it. Other peers in the channel fetch the block using its path. The peer node then requests to join the channel, which is approved if the node is configured for the requested channel. Otherwise, the request is rejected. The process is detailed in Fig. 7.

Fig. 7. Joining Channels

Industry 4.0 Smart Supply Chain Data Protection Using Block Chain

6

205

Conclusion

The article discusses the use of blockchain technology in finance and supply chain management and proposes a design for a blockchain system based on Hyperledger Fabric. The system’s security and confidentiality features are discussed, and results show the system’s success in a prototype phase, demonstrating its applicability outside of the finance system in a cost-effective way. However, the article notes that the system lacks the automation property of generating network artifacts as it is not deployed on the network, and all testing was done locally using scripts. The network should be deployed on the cloud, which would increase system automation, to increase efficiency. However, the incorporation of cloud computing and other configurations would also lead to an increase in system complexity. Involving more network participants is advised, in addition to those covered in the network design chapter, to increase scalability. By switching from the Solo consensus algorithm to the Kafka consensus algorithm, which calls for the use of more orderer nodes and does away with the ordering system’s single point of failure, the system could be made more suitable for use in production. Nevertheless, adding more orderer nodes and the additional configurations needed for the Kafka algorithm would make setting up the system more difficult.

References 1. Dittmar, R., Andrew Wu, F.: Blockchain and Cryptocurrency Explained. Michigan University. www.coursera.org/learn/crypto-finance 2. Parizo, C.: What are the 4 different types of blockchain technology? www. searchcio.techtarget.com/feature/What-are-the-4-different-types-of-blockchaintechnology. Accessed May 2021 3. Iredale, G.: What are the Different Types of Blockchain Technology? www. 101blockchains.com/types-of-blockchain/. Accessed Jan. 2021 4. Geroni, D.: Hybrid Blockchain: The Best of Both Worlds. www.101blockchains. com/hybrid-blockchain/. Accessed Jan. 2021 5. gRPC vs REST: Understanding gRPC, OpenAPI, and REST and When to Use them in API Design. www.cloud.google.com/blog/products/api-management/ understanding-grpc-openapi-and-rest-and-when-to-use-them. Accessed Apr. 2020 6. Kamath, R.: Food traceability on blockchain: Walmart’s pork and mango pilots with IBM. J. Br. Blockchain Assoc. https://doi.org/10.31585/jbba-1-1-(10)2018 7. Hyperledger-fabricdocs, Hyperledger Fabric Documentation. www.hyperledgerfabric.readthedocs.io/en/latest/deployorderer/ordererplan.html#:∼:text=In %20a%20Hyperledger%20Fabric%20network,and%20commit%20to%20their %20ledgers. Accessed Dec. 2022 8. PROOF-OF-STAKE (POS). www.ethereum.org/en/developers/docs/consensusmechanisms/pos/. Accessed Mar. 2023 9. What is Gossip Protocol? John Fornell. www.academy.bit2me.com/en/ que-es-gossip-protocol/#:∼:text=In%20blockchain%20networks%2C%20this %20protocol,is%20spread%20through%20social%20networks. Accessed Jun. 2020

206

M. A. Abu El Magd et al.

10. Chang, S.E., Chen, Y.: When blockchain meets supply chain: a systematic literature review on current development and potential applications. Ieee Access 8, 62478–62494. https://doi.org/10.1109/ACCESS.2020.2983601 11. Natanelov, V., Cao, S., Foth, M., Dulleck, U.: Blockchain smart contracts for supply chain finance: mapping the innovation potential in Australia-China beef supply chains, vol. 30. https://doi.org/10.1016/j.jii.2022.100389 12. Raj, P.V.R.P., Jauhar, S.K., Ramkumar, M., Pratap, S.: Procurement, traceability and advance cash credit payment transactions in supply chain using blockchain smart contracts. Comput. Ind. Eng. 167, 108038vol (2022). https://doi.org/10. 1016/j.cie.2022.108038

Methodological Model for the Solution of Periodic Customer Scheduling in Routing Problems Restrepo Franco Alejandra Mar´ıa1,3,4(B) , Valencia Rodriguez Orlando1,5 , Toro Ocampo Eliana Mirledy6 , Bravo Ort´ız Mario Alejandro2,3,4 , Cardona Ramirez Nicolas4 , Orjuela Paez Cristian Camilo3 , and Valencia D´ıaz Mario Andr´es3 1

5

Departamento de F´ısica y Matem´ aticas, Universidad Aut´ onoma de Manizales, Manizales, Colombia [email protected] 2 Departamento de Electr´ onica y Automatizaci´ on, Universidad Aut´ onoma de Manizales, Antigua Estaci´ on del Ferrocarril, Manizales CP 170001, Colombia [email protected] 3 Sigma Ingenieria, Manizales Caldas, Colombia 4 Geostrategy, Manizales Caldas, Colombia [email protected],[email protected] Departamento de Ingenier´ıa Industrial, Universidad Nacional de Colombia, Sede Manizales, Colombia 6 Ingenier´ıa Industrial, Universidad Tecnol´ ogica de Pereira, Pereira, Risaralda, Colombia [email protected]

Abstract. Transportation logistics represents a challenge in the efficient management of supply chains of products and services in urban centers that require effective solutions to properly manage operations, one of these points is linked to the creation of routes based on the demand of users considering geographical conditions, size, and capacity of the fleet, points that have the opportunity to be integrated into the Vehicle Routes theme. This initiative recommends a methodological model to solve the periodic scheduling of consumers associated with the PVRP, where the service is defined by the frequency of visits to the nodes; to be implemented in an information technology (IT) industry for logistics management in the city of Manizales—SIGMA Ingenier´ıa—to validate the development of a logistics platform based on step-by-step consumer scheduling.

Keywords: Vehicle Routing Problems Clusters

· Periodical Scheduling ·

c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 207–218, 2023. https://doi.org/10.1007/978-3-031-36957-5_18

208

1

R. F. Alejandra Mar´ıa et al.

Introduction

Logistics is a challenge in the efficient management of supply chains of products and services in urban centers that require effective solutions in order to adequately manage aspects such as vehicular congestion generated by the high population density, timely attention in the location of customers, periodic attention in the case of services such as garbage collection, attention to contingencies such as failures in the continuity of energy, water and internet services, compliance with differentiating policies such as minimum service times, among others [1]. One of these aspects is associated with the preparation of routes based on user demand, geographic conditions, size and capacity of the vehicle fleet, aspects that can be included in the subject called vehicle routing, on which this proposal is based. The vehicle routing problem (VRP) is one of the most attractive topics in operations research, logistics and supply chain management. VRP is concerned with minimizing the total cost of logistics systems [2], which in turn are wellknown combinatorial optimization problems that arise in the areas of transportation, usually involving scheduling in constrained environments. In transportation management, there is an obligation to provide services from one supply point (depot) to several geographically dispersed points (customers) with significant economic implications. Due to the important applications of VRP, many researchers have developed solution approaches for such problems [3] and although it has been studied for almost six decades [4], actual applications remain challenging as they present a variety of operational attributes that complicate the problem and can have a significant impact on the solution process. Examples can be found in grocery (or other product) delivery, waste collection, or equipment distribution for intermodal operations [5]. In these examples, routes must be designed to account for volume, travel time, and frequency of visits. The periodic vehicle routing problem (PRVP) is a generalization of the classical vehicle routing problem (VRP), in which vehicle routes are constructed for a period of t days (e.g., one week). However, other units of time can be used. The objective of the VRP is to find a set of routes for each vehicle over the period that minimizes total travel time while satisfying operational constraints (vehicle capacity and visit requirements) [6]. The above demonstrates that transportation decisions in modern companies are made in the context of integrated supply chains. The tactical and operational levels of transportation comprise medium- and short-term decisions, including detailed planning of visit schedules, routes, and loading plans. The right synergy between decision levels helps to consolidate the supply chain and is a recurring challenge for planners. The routing problems and their different variants of the solution, although they have been great contributions to the literature, the challenge has been to integrate them into systems or software platforms, showing one of the most evident gaps between research and industry [7], although the developments based on mobility are clear and accurate, the issue of creating and scheduling the

Methodological Model for the Solution

209

route with different customers and conditions becomes a challenge for companies engaged in transportation in different sectors. Therefore, this article focuses on developing a methodological model for the solution of the periodic scheduling of customers articulated to the PVRP, where the service is defined by the frequency of visits to the node; which will be implemented in a logistics management Information Technology (IT) industry in the city of Manizales—SIGMA Ingenier´ıa, to validate the step by step when developing a logistics platform based on customer schedules. The objective of this proposal is to propose a solution methodology that serves a set of customers efficiently using a fleet of available vehicles which is discussed in the following sections: Related works to identify solution alternatives, problem formulation, proposed methodology, results, and conclusions.

2

Related Works

Vehicle routing is of great importance in the literature and has been extensively studied due to its relevance to the industry. The first problem that motivated PVRP was introduced by Beltrami and Bodin [7] for allocating lift compactor trucks in municipal waste collection. PVRP was formally defined by Russell and Igo [?] as a “route assignment problem” and was first mathematically formulated by Christofides and Beasley [8]. The solution methods in these and subsequent works have focused on two-stage heuristics (construction and improvement); such a study started with the VRP proposed by Dantzig and Ramser [3] where vehicles with the same characteristics leave a distribution center (or warehouse) and deliver or pick up goods at specific locations (which can be customers or suppliers) at the lowest possible cost. The cost in this case is associated with the distance traveled. Gaudioso and Paletta [3], consider a PVRP model that minimizes fleet size, Cordeau, Gendreau and Laporte [9] apply a tabu search algorithm for PVRP. All these works consider heuristic methods whose quality is unknown because neither optimal solutions nor lower bounds are provided. However, since these works share common data sets, the results are compared between them. In the aforementioned references, each client is visited at a prespecified frequency. Each node can be served from a specific set of schedule options with a fixed number of visits per week. But it is important to understand, that over time, a paradigm shift in urban logistics services has been generated as the demand for instant and real-time delivery and mobility services grows. This poses new challenges for logistics service providers, as the underlying dynamic vehicle routing problems require anticipatory real-time routing actions [12]. Real-world VRP has become increasingly dynamic, and global interconnectedness, urbanization, ubiquitous information flows, and greater service orientation have only increased the dynamism and urgency of service fulfillment [12]. New services, such as same-day or restaurant meal delivery, shuttle service, and emergency repairs, force logistics service providers to anticipate future demand,

210

R. F. Alejandra Mar´ıa et al.

adjust to real-time traffic information or even incorporate unknown drivers of collaboration to meet customer expectations [13]. Dynamism adds another dimension to vehicle routing optimization. The problem is not only finding suitable routing in a large and constrained space but also evaluating routing actions with respect to future dynamic developments [14]. One of these techniques is clustering, which has been used to solve different problems applied in different fields. The K-means algorithm and its various extensions are popular and commonly used in various clustering studies [15]. It was proposed by MacQueen [16]. The algorithm is an unsupervised machinelearning method used in data mining and pattern recognition. The K-means algorithm has the advantages of brevity, efficiency, and speed. Minimizing the sum of squared errors, which is the objective function, is the basis of this algorithm [17]. K-means of [18] is one of the most popular clustering methods. In Algorithm 1. The algorithm shows the K-means clustering procedure. The basic idea is: given an initial but non-optimal clustering, relocate each point to its nearest new center, update the clustering centers by calculating the mean of the member points, and repeat the relocation and updating process until the convergence criteria (such as a predefined number of iterations, the difference in the value of the distortion function) are met. The initialization task is to form the initial K clusters [18]. And on the other hand, in [19] analyzed methods for clustering that can be used to deal with spatial and temporal patterns in a large amount of data. Their approach allowed observing the existence of different spatial and temporal families. On the other hand, in [20] proposed the K-means Algorithm for the problem of increasing data, generated dynamically and without repetition, which reduces the computational time, providing more accurate results. So the initial clustering was performed on the statistical data, using K means. One of the most common problems for the situations presented is the Traveling Agent Problem (TAP), which is the general situation and starting point for formulating other more complex, but more practical, combinatorial problems, such as vehicle routing and enlistment time-dependent task scheduling, this process has been addressed by means of clustering and genetic algorithms by Rani Algorithm 1 K-means clustering algorithm [18] Require: K, number of clusters; D, a dataset of N points Ensure: A set of K clusters 1: Initialization 2: repeat 3: for each point p in D do 4: find the nearest center and assign p to the corresponding cluster 5: end for 6: update clusters by calculating new centers using mean of the members 7: until stop-iteration criteria satisfied 8: return clustering result

Methodological Model for the Solution

211

et al. [21], the algorithm was used to solve an itinerary problem for a tourist, the results demonstrate the feasibility of this approach to solve problems of this nature or which we henceforth refer to as TAP in addition to being considered as one of the most difficult problems to solve, however, it has been very useful to solve logistics problems in the industry. The TAP consists of finding the shortest route by visiting n cities on a single occasion and returning to the city of origin, a mathematical model based on this study [22]. The methods and techniques mentioned above have been used with the intention of grouping a set of data, which are part of a problem to be solved, with the intention of reducing its computational complexity.

3

Problem Formulation

SIGMA Ingenier´ıa builds smart cities through Geographic Information Systems as technological tools for the analysis of large volumes of information and support to the strategic processes of customers, specifically public utilities of sanitation, environment, and energy. Each of its products provides specialized support in its business model for data management, maintaining data availability and integrity throughout the value chain. One of its systems is logistics software that manages transportation operations in order to maintain control of the information generated from the operations carried out by transportation systems. One of the difficulties that have arisen is to develop a methodology that leads the system to schedule optimal routes to different customers with varying frequencies of visits in a range of one month, complying with different constraints such as: – Customers must be visited with a given frequency. – 9 types of visit frequency to be executed in a month must be complied with. – The vehicle fleet must start and end its run at the central depot and a truck can visit several customers on the same day if it reaches capacity. – Trucks are available every day of the week, with the premise that on Sunday only 14% of the total fleet (2 Vehicles) can operate. Although the trucks have limited capacity, the total available travel time is assumed to be the dominant constraint, hence frequencies become relevant target variables. – In terms of time, the following should be taken into account: the average duration of service for each customer, the customer service schedules, and the shift established for the operation, which would be two shifts, the first shift shall begin at 7:00 AM or 8:00 AM and the second shift shall be worked between the hours of 2:00 PM and 7:00 PM, a maximum of 5 h each. For the solution of the problem it was determined to start from the historical information of a logistic operation of transport which is composed of 2660 clients that are in a metropolitan area, this in turn has the variables of frequency of collection, schedules of attention of the client, average time of attention in point, kilogram/frequency of demand and geographic location. The operation has 14 vehicles available but 2 must be kept in stock.

212

3.1

R. F. Alejandra Mar´ıa et al.

Variable Definition

Input parameters – – – – – –

Ci: Customer, it is categorized with the location in Latitude and Longitude. IT: work schedules established by each Ci. ti: service time at node i. Ti: Time window that defines the vehicle’s working day. k: vehicles. kf: average kilogram/frequency history for each Ci.

Decision variables Frequencies (fi): Decision variable in the scheduling model Types of fi: – fi, i = 1 visit to the client once a month. – fi, i = 2 visits the client twice a month (e.g. in the first and third weeks or in the second and fourth weeks). – fi, i = 4 visit the customer four times a month (visit the customer 1 time each week). – fi, i = 8 visit the client twice every week. – fi, i = 12 visit the client three times each week. – fi, i = 16 visit customer four times a week. – fi, i = 20 visits customer five times each week. – fi, i = 24 visit customer six times a week. – fi, i = 30 visit the customer every day of the week. Variable to optimize dij : Node distance One of the most widely used algorithms to reduce the distance between geographic points is K-Means [?], which is a totally accurate method that contributes to the subdivision of the territory into small zones that allow vehicles to reduce their travel time. The k-means is one of the most popular unsupervised learning algorithms that solve the well-known clustering problem and its mathematical contribution is described below. Let X = {x1 , ...., xn } be a data set in a d-dimensional Euclidean space Rd . Let A = {a1 , ...., an } be the c cluster centers. Let Z = [Zij ]nxc , where Zij is a binary variable (i.e Zij ∈ {0, 1}) indicating if the data point xi belongs to k-th cluster, k = 1,..., c. The k-means objective function is: n c J(z, A) = i=1 k=1 Zik  xi − ak 2 . The k-means algorithm is iterated through necessary conditions for minimizing the k-means objective function J(z, A) with updating equations for cluster centers and memberships, respectively, as:

ak =

n i=1 Zik Xij  . and n i=1 Zik

Methodological Model for the Solution

 Zik =

2

213

2

1, if xi − ak  = min1≤k≤c xi − ak  otherwise .

0,

where xi − ak  is the Euclidean distance between the data point xi and the cluster center ak . For the application of this paper, X is Latitude and Longitude and A are the optimized centroids. Now the objective is to find the shortest cyclic path that passes through all the given nodes. A generalization of this problem is the TSP problem. A set of nodes is divided into clusters and only one node of each cluster must be visited and make a cyclic tour returning to the initial node/cluster. Under this premise, the mobility solution is given the different routes are generated, and their mathematical contribution [23] is presented below: Minimise total cost: min

 i

Cij xij

j

Enter each city once: 

xij = 1 for all j

i

Leave each city once:  xij = 1 for i=2,...,ni = j, j = 2, ..., n j

Subtour breaking: ui + 1 ≤ uj + n(1 − xij ), xij  {0, 1}

4

for i=2,...,ni = j, j = 2, ..., n,

for all i,j; ui  0 for all i

Methodology

To solve the combined visit scheduling and routing problems, 3 solution approaches are introduced. First, to minimize the objective function of the proposed model, an algorithm was developed, which initially solved the distance reduction between nodes using an exact approach based on K-means, then the scheduling challenge was solved using an algorithm designed from the historical information of scheduled visits to customers based on the restrictions established in the formulation of the problem. Finally, the solutions were articulated by balancing loads of the sectors and entering the package to the mobility model of the company and generating the different routes that satisfy 100% of the transportation operation.

214

5 5.1

R. F. Alejandra Mar´ıa et al.

Results Node Distance Reduction (dij )

In order to reduce the distance between nodes, the K-means clustering algorithm explained in Sect. 2 was applied. For the K-means model, the input variables (X) were the latitude and longitude of each client, seeking to create optimal clusters in the sector to program the routes, for this, the elbow method is applied, which uses the average distance of the observations to their centroid. In other words, the intra-cluster distances are considered. The larger the number of clusters k, the intra-cluster variance tends to decrease. The smaller the better, as it means that the clusters are more compact. Initially, there are 2604 customers, which are clustered by distance and as shown in Fig. 1, to be more compact clusters, the most optimal number is three, the point where the curve begins to stabilize. The red cluster had 95 points, the green cluster had 499 points and finally, the purple cluster had 2102 points, which is too high to balance the scheduling of the routes (Fig. 2).

Fig. 1. Initial clustering of spatialized points

Fig. 2. Spatialization of the information newly identified clusters

According to the results obtained, clustering was performed again for cluster 3 (purple) in order to further disaggregate the points and reduce the complexity of the programming taking into account the constraints. As shown in Fig. 3, the

Methodological Model for the Solution

215

optimum indicates 4 clusters. Next, the geographical distribution of the clients is presented, which shows the density of visits and the spatialization of the information with the newly identified clusters. 5.2

Solving the Visit Scheduling Problem

Taking into account the distribution of customers, the periodic scheduling design is carried out with a total of 6 clusters identified in the previous section. In the visit scheduling model, the optimal decisions depend on the control parameters of each point, in this case the frequency, the time windows of each customer and the average time used in point. To provide an approximate result that is comparatively simple to implement for a system, the heuristic dynamic scheduling is proposed, whose basic principles are described in Fig. 3, which is represented under the dynamics of the frequencies. A new clustering algorithm is then generated where the input variable is the frequency, thus generating that customers with frequency 1 have 28 clusters, i.e. they can be divided into 28 days of the 4 weeks of the month. For frequency 2, it is partitioned into 14 clusters, i.e. 14 days visiting customers every 2 weeks. For frequency 4, it is divided into 6 clusters, i.e. visiting once a week; for frequency 8, 3 clusters are generated to visit twice a week; for frequency 12, 2 clusters are generated to visit 3 times a week. For frequency 16, as it is visited 4 times a week, the points are assigned from Wednesday to Saturday taking the decision of the distribution by capacity limits to be collected, similarly, frequency 20 is assigned from Monday to Friday, frequency 24 from Monday to Saturday and frequency 30 from Monday to Sunday. For this process, two types of calendars were designed, the Ordinary, which is designed to accommodate between 150 and 1500 clients, and the Special, which

Fig. 3. Initial clustering of spatialized points.

216

R. F. Alejandra Mar´ıa et al.

is conditioned for a cluster with a maximum of 150 clients who do not demand more than frequency 8 in the clusters. The established working hours are from 8:00 am to 12:00 noon and 2:00 pm to 8:00 pm and as a result of the process 6 clusters were generated, which are described below: – Cluster 1–0 grouped 1047 clients, using 2 vehicles of 1200 Effective Capacity (EC) (Kg), programmed in two days which travel 76 routes. – Cluster 1–1 grouped 469 clients, using 200, 500, and 900 EC vehicles, in double shifts running 45 routes. – Cluster 1–2 grouped 395 customers, using 200 and 300 vehicles, in double shifts traveling 39 routes. – Cluster 1–3 with 493 customers using 500 and 1200 CE vehicles in one day traveling 31 routes. – Cluster 1–4 with 95 customers uses only one 500 CE vehicle in a day, which runs 12 routes. – Cluster 1–5 with 102 customers runs 8 routes in a day with one 200 CE vehicle.

6

Conclusions

According to the above, and the results of the algorithms implemented, for the total number of customers we were able to collect approximately 2005 kg of load per day, in days that had an average time of 4.5 h without exceeding 100 km per route, which were trips to other cities. On the other hand, 4 vehicles were kept in stock, thus generating a fleet optimization of 28.8%. With this, it is possible to find feasible solutions for scheduling routes to customer visits through a new way of grouping the problem into clusters with the intention of creating subproblems, which are solved by clustering algorithms by distance. In the application of the designed methodology to the test case, the automation process takes 1–2 min per cluster to balance the routes of the 2604 customers. This makes it an efficient model for changing operations and increasing amounts of customers/visits over time. The methodological model allows to have the step-by-step process of scheduling customer visits taking into account the established constraints, also the proposal of this paper focuses on the combination of the use of k-means with the grouping of routing scheduling elements, by innovating a way of grouping without generating a high computational complexity, which is proven derived from the simple operations required for its calculation. The results can lead to the conclusion that this alternative could be very convenient in practice if medium-level processing is available. Consequently, it is also deduced that the general objective of developing an option to improve the planning and assignment of transportation routes in companies is fully accomplished.

Methodological Model for the Solution

217

References 1. Arvianto, A., Sopha, B.M., Asih, A.M.S., Imron, M.A.: City logistics challenges and innovative solutions in developed and developing economies: a systematic literature review. Int. J. Eng. Bus. Manag. 13 (2021). https://doi.org/10.1177/ 18479790211039723 2. Zhou, W., Song, T., He, F., Liu, X.: Multiobjective vehicle routing problem with route balance based on genetic algorithm. Discrete Dyn. Nat. Soc. (2013). https:// doi.org/10.1155/2013/325686 3. Vieira, B.S., Ribeiro, G.M., Bahiense, L.: Metaheuristics with variable diversity control and neighborhood search for the Heterogeneous site-dependent multi-depot multi-trip periodic vehicle routing problem. Comput. Oper. Res. 153, 106189 (2023). https://doi.org/10.1016/J.COR.2023.106189 4. Scheinberg, K., et al.: MOS-SIAM Series on Optimization (2014). Accessed 01 Mar 2023. www.ibm.com 5. Mindur, L., Mindur, M.: Intermodal transport on selected continents. Stud. Syst. Decis. Control 400, 125–195 (2022). https://doi.org/10.1007/978-3-030-87120 6. Francis, P., Smilowitz, K., Tzur, M.: The period vehicle routing problem with service choice. 40(4), 439–454 (2006). https://doi.org/10.1287/TRSC.1050.0140 7. Beltrami, E.J., Bodin, L.D.: Networks and vehicle routing for municipal waste collection. Networks 4(1), 65–94 (1974). https://doi.org/10.1002/NET.3230040106 8. Christofides, N., Beasley, J.E.: The period routing problem. Networks 14(2), 237– 256 (1984). https://doi.org/10.1002/NET.3230140205 9. Gaudioso, M., Paletta, G.: A heuristic for the periodic vehicle routing problem. Transp. Sci. 26(2), 86–92 (1992). https://doi.org/10.1287/TRSC.26.2.86 10. Cordeau, et al.: VRP-REP: the vehicle routing problem repository (1997). www. vrp-rep.org/references/item/cordeau-et-al-1997.html. Accessed 01 Mar 2023 11. Chao, I.-M., Golden, B.L., Wasil, E.: An improved heuristic for the period vehicle routing problem. Networks 26(1), 25–44 (1995). https://doi.org/10.1002/NET. 3230260104 12. Hildebrandt, F.D., Thomas, B.W., Ulmer, M.W.: Opportunities for reinforcement learning in stochastic dynamic vehicle routing. Comput. Oper. Res. 150, 106071 (2023). https://doi.org/10.1016/J.COR.2022.106071 13. Ulmer, M.W.: Dynamic pricing and routing for same-day delivery. 54(4), 1016– 1033 (2020). https://doi.org/10.1287/TRSC.2019.0958 14. Bagloee, S.A., Tavana, M., Asadi, M., Oliver, T.: Autonomous vehicles: challenges, opportunities, and future implications for transportation policies. J. Modern Transp. 24(4), 284–303 (2016). https://doi.org/10.1007/S40534-016-0117-3/ FIGURES/5 15. Ikotun, A.M., Ezugwu, A.E., Abualigah, L., Abuhaija, B., Heming, J.: K-means clustering algorithms: a comprehensive review, variants analysis, and advances in the era of big data. Inf. Sci. (N Y) 622, 178–210 (2023). https://doi.org/10.1016/ J.INS.2022.11.139 ¨ ¨ urk, G., Ozer, ¨ 16. Ay, M., Ozbakir, L., Kulluk, S., G¨ ulmez, B., Ozt¨ S.: FC-K means: Fixed-centered K-means algorithm. Expert Syst. Appl. 211, 118656 (2023). https://doi.org/10.1016/J.ESWA.2022.118656 17. Arthur, D., Vassilvitskii, S.: K-means++: the advantages of careful seeding 18. Mannor, S., et al.: K-means clustering. In: Encyclopedia of Machine Learning, pp. 563–564 (2011). https://doi.org/10.1007/978-0-387-30164-8425

218

R. F. Alejandra Mar´ıa et al.

19. Nagy, M., Negru, D.: Using clustering software for exploring spatial and temporal patterns in non-communicable diseases. Euro. Sci. J. ESJ 10(33), 1857–7881 (2014). https://doi.org/10.19044/ESJ.2014.V10N33P 20. Byun, J.W., Kamra, A., Bertino, E., Li, N.: Efficient k-anonymization using clustering techniques. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4443, pp. 188–200. LNCS (2007). https://doi.org/10.1007/978-3-540-71703-418 21. Rani, S., Kholidah, K.N., Huda, S.N.: A development of travel itinerary planning application using traveling salesman problem and k-means clustering approach. In: ACM International Conference Proceeding Series, pp. 327–331 (2018). https://doi. org/10.1145/3185089.3185142 22. Fuentes, G.E.A., Gress, E.S.H., Mora, J.C.S.T., Mar´ın, J.M.: Soluci´ on al Problema de Secuenciaci´ on de Trabajos mediante el Problema del Agente Viajero. RIAI Revista Iberoamericana de Automatica e Informatica Industrial 13(4), 430–437 (2016). https://doi.org/10.1016/J.RIAI.2016.07.003 23. el Krari, M., Ahiod, B., el Benani, Y.B.: A pre-processing reduction method for the generalized travelling salesman problem. Oper. Res. 21(4), 2543–2591 (2021). https://doi.org/10.1007/S12351-019-00533-W/TABLES/10

Development of an Information System for the Monitoring of Physiological Variables and Information Storage of Neonates on Domiciliary Oxygen J. L. Amaya-Carrascal1 , C. A. Vera-Betancourt1 , C. Marquez-Narvaez1 , S. Murillo-Rend´ on1,2(B) , I. Echeverri-Ocampo1 , D. Henao-D´ıaz1 , Belarmino Segura-Giraldo3 , and C. Salgado-Jim´enez3 1

3

Universidad Aut´ onoma de Manizales, Manizales, Caldas, Colombia [email protected] 2 Universidad de Caldas, Manizales, Caldas, Colombia Universidad Nacional de Colombia, sede Manizales, Caldas, Colombia

Abstract. The development of information systems has increased over the years, and the health area is no exception given the interest of professionals in the use of information systems that help make decisions based on the data taken. An information system for the management and monitoring of information related to neonates on home oxygen was developed using the Scrum framework and the incremental iterative development model, in addition to different web development technologies. The development of an information system with 12 modules that allow the monitoring of physiological variables of oxygen-dependent neonates as well as the management of related information was achieved. The resulting tool offers important functionalities, including the possibility of monitoring vital signs and issuing alerts based on the values of these signs. For the follow-up in the development process of each of the modules of the information system, the Scrum framework was used, using the incremental iterative software development model. A tool was developed through the integration of development methodologies, languages, and programming techniques that supports the process of health professionals in institutions that monitor the evolution of neonates with home oxygen through the use of telecare.

Keywords: Neonates

· Information system · Telecare

Supported by MINCIENCIAS financiacin Cdigo: 121984468173, Contrato No. 4142020, Universidad Autnoma de Manizales, Universidad Nacional de Colombia, SES Hospital Universitario de Caldas, Clnica Ospedale. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 219–230, 2023. https://doi.org/10.1007/978-3-031-36957-5_19

220

1

J. L. Amaya-Carrascal et al.

Introduction

Information systems are computer-supported systems that provide their users with specific information of interest for a given organizational context [10]. They are responsible for systematically maintaining information for later use and analysis [16]. Information systems have many areas of application within society. the Internet itself has been developed based on large-scale information systems, which are critical for the execution of personal and business operations [17]. The development of information systems in different areas has increased over the years. Thus, the applications of information systems are seen in various fields, from education, where they are used to keep track of students’ grades and other institutional data [1], to business, where they are used to store, consult and perform important monetary transactions [11]; also in agriculture, where they are used to train future farmers; in sports, to manage the sale of sports tickets [4]; and in the area of electronic commerce, to manage the products offered [12]. Additionally, this increased interest in the development of information systems is evidenced by the work of experts and researchers in the health sector [9]. Literature review [8] shows that the development of health information systems represents a unique opportunity to learn, create and expand the work in the area of this type of system. The following is some of the background information related to the development and implementation of information systems in different fields associated with the area of health at the international and national levels. In the international context, in the United States, some scientists have shown great interest in research supported by information systems for decision-making and data capture. For example, particularly in children, researchers from the Journal of the American College of Cardiology have relied on the Pediatric Health Information System, which contains data from 2004 to 2015 on 44 nationally distributed pediatric hospitals, to conduct a study about the management of newborns with supraventricular tachycardia (SVT). As a result, they obtained that the use of propranolol in the newborn with SVT is associated with lower mortality and hospital costs compared to digoxin [6]. In another study conducted in the United States to identify risk factors for catastrophic adverse outcomes in children with pulmonary hypertension (PH) undergoing cardiac catheterization, using the National Pediatric Health Information System and data stored in it during the period from 2007 to 2012, they obtained as a result that the risk of cardiac catheterization in children and young adults with PH is high compared to the risk previously reported in other pediatric populations. It was also found that this risk is influenced by different individual patient factors [13]. According to the documentation reviewed, referring to the national level, a scarcity is denoted regarding the use of health information systems for the care of children and/or neonates, however, Bernal, in his study [5], where used a conceptual framework, However, Bernal, in his study [5], in which he used a conceptual framework, interviews, and literature review as a methodology, managed to determine that Colombian health information systems have major quality problems and therefore suggests taking advantage of the change that the health system is undergoing

Development of an Information System for the Monitoring . . .

221

to improve the information; in addition, he points out that simplification and standardization of the information capture mechanisms are necessary. Taking into account all of the above, this paper proposes the development of an information system in the research project “Telecare for home oxygen monitoring in neonates in Manizales” for the management (record, query, update and delete) of the information of neonatal patients in telecare with home oxygen, their medical history and that of their caregiver. In addition, the system includes the monitoring and follow-up of the physiological variables of the neonates, the generation of reports based on the information stored and the sending of alerts related to the vital signs of the neonates. The need to develop this information system arises since the current tools or systems found do not offer the possibility of real-time monitoring of the physiological variables of the users, which is the general purpose of this tool, in addition to the support it has with IoT. Added to the above is the fact that the aim was to have the possibility of recording and storing the clinical history, background, and all the information related to neonatal patients and their parents; all this for the sake of its subsequent study by researchers, physicians of the allied entities and other users. This is why it was decided to develop such a tool with which it will be possible to have total control over the form and the servers in which the information is stored. In addition, it will be possible to define and decide who has access to certain data or functionalities according to the roles and permissions designed; likewise, it will be possible to develop the functionalities that allow the telecare devices to connect to the information system to load the vital data collected from the patients.

2

Methodology

This study focused on the development of an information system that works as a web application. This application includes different modules that allow the management and monitoring of information related to neonates with respiratory problems. The development of the project was divided into two parts: software development and platform development. 2.1

Software Development

During the development of the project, the agile development methodology called Scrum was used. This methodology offers a compilation of phases that allow the application of good practices in a project throughout its life cycle [15]. The phases that make up the Scrum methodology can be seen in Figure 1. The project activities were divided into different iterations that were developed in a certain period. Each of these iterations in the Scrum methodology is known as a sprint. That is, a sprint is a fixed period in which a series of established tasks are carried out that result in a new increment in the functionality of the final system [14]. For the project, each sprint had a duration of four weeks and in each sprint, the tasks to be executed were prioritized.

222

J. L. Amaya-Carrascal et al.

Fig. 1. Scrum methodology.

To follow up on the progress of each of the sprints, a daily meeting known as a scrum daily meeting was held. Each meeting, with a maximum duration of fifteen minutes, sought to evaluate the progress in the activities of each team member, as well as to know the difficulties that could arise. Additionally, the product backlog and sprint backlogs were performed in the Jira platform [2], which is an online tool for managing project tasks. The agile methodology chosen was worked together with the incremental iterative software development model. With the iterations carried out, increases in the functionality of the information system were achieved, reaching, with each one of them, the development of a new module of the platform. Additionally, with each iteration, the phases of the software development life cycle were fulfilled. Thus, the analysis and design phases resulted in class diagrams and entity-relationship diagrams. On the other hand, for the implementation phase, the Laravel development framework [3], which is based on the PHP programming language, was mainly used. This framework offers the possibility of developing a complete web application since it has integrated tools such as Blade, which is a template engine that, together with CSS and JavaScript, allows the development of the front end of an application. Likewise, Laravel is considered a framework especially focused on the back end of an application, which is why it has different mechanisms that allow the implementation of the platform’s business logic.

Development of an Information System for the Monitoring . . .

223

Regarding the development model, for this information system, each existing module corresponds to an iteration of the model used, thus achieving, with each iteration, an increase in the final product, but, at the same time, corrections and modifications of previously developed modules. 2.2

Platform Development

The web platform is composed of twelve modules that allow for storing a large amount of information related to the neonates that are registered in the information system. Among these modules are the initial module, mothers, fathers, neonates, devices, oxygen characterization, neonatal controls, alerts, discharges, users, roles, and permissions Table 1. System security and data access are controlled based on the roles and permissions associated with each user that enters. Thus, depending on the privileges associated with it, each module can be managed completely or only the visualization of the information about it. The existing roles in the system can be managed at any time by an Administrator user, which means that there can be as many roles as the system administrator wishes, considering that these roles are associated with specific system permissions, when creating or editing a role, it is necessary to select which permissions will be available for that role. The information system has 62 permissions that reflect the actions that a user can perform within the system and that should be controlled so that not just any user can perform them. Communication between the different parts of the system is given, for the most part, by the use of the HTTP protocol, but it should be clarified that sockets are also used under the TCP protocol for real-time communication of the physiological variables of the neonates. For this communication, use is made of a specific port on the server where the web application is hosted; the architecture used for the development of the information system can be seen in Fig. 2. 2.3

Results

The results obtained in the project are presented in terms of the two components proposed in the methodology. On the one hand, the results are presented concerning the software development and, on the other hand, the results of the platform development. Software Development The results obtained in terms of software development have been aligned with the selected development methodology (Scrum). Therefore, during a period of 12 months with sprints of one month, all the phases defined in the framework were followed, which made it possible to On the other hand, multiple modules were developed in the system, which was planned as iterations, according to the iterative and incremental software development model. Each module developed meant an increase in the final product to be obtained, and during the development of a module, corrections, and modifications were also made to the modules already developed. Thus, the development was iterative.

224

J. L. Amaya-Carrascal et al.

Table 1. Description of information system modules. Module

Description

Intial

The initial module allows the monitoring of the neonates’ vital signs, as well as the visualization of the alerts issued by the information system

Mothers

A module that allows the management of the information of the mothers of the newborns registered in the system.

Parents

A module that allows the management of information related to the parents of the newborns registered in the system.

Neonates

This module allows the management of neonate information, i.e., adding, viewing, updating, and deleting records. In addition, it offers the possibility of viewing reports of measurements and alerts generated by each of the neonates registered on the platform

Devices

This module allows the management of the devices installed in the neonates’ homes so that each neonate can be linked to a module that sends information on the newborn’s physiological variables.

Oxygen Characterization The oxygen characterization module offers the possibility to manage information on home oxygen administered to neonates Neonatal Checkups

A module that allows managing the information of the medical controls that are performed on the newborns registered in the information system

Alerts

Unlike the other modules, the alerts module only allows the visualization of alerts that are generated from neonatal vital sign measurements, providing the user with information about what type of alert occurred and when it was triggered

Expenses

Module in charge of managing the information of the discharges that are carried out in the health institution, allowing a follow-up in the evolution of the newborn

Users

This module allows system administrators to manage the users that have access to the platform, offering the possibility of enabling or disabling these users when the administrator considers it necessary

Roles

The roles module allows the administrator user(s) to manage the roles that will later be linked to the users registered in the system, so, when deemed necessary, a user with the Administrator role can create, delete or edit the existing roles in the platform

Permits

As with the alerts module, the permissions module only allows the visualization of information, in this case, such information will be the permissions registered in the platform and that are part of the roles. These permissions can be reflected as actions that are allowed to be performed within the information system

Development of an Information System for the Monitoring . . .

225

Fig. 2. Information system architecture.

It is worth mentioning that the development of each of the modules consisted of the execution of different activities, such as analysis, design, implementation, and testing, which are necessary tasks for the correct development of software. Finally, the modules developed in the information system were: initial module, mothers, fathers, neonates, devices, oxygen characterization, neonatal controls, alerts, discharges, users, roles, and permissions. Platform Development The web platform was developed using client/server architecture; this web application was developed using the Laravel programming framework, version 8.43.0. The developed application contains both the back end and the front end. Therefore, on the one hand, there is the business logic and data access and, on the other hand, there is the data presentation layer and the interaction with the system users. Additionally, the back end of the application has a security and data access system based on roles and permissions, which offers the possibility of managing the roles that will be subsequently assigned to users. These roles allow limiting the actions that an authenticated user can perform. Finally, for the management of roles and permissions of the system, the laravel/permissions package was used; this is available for Laravel and is developed and maintained by the Spatie community. For the front end of the application, we used different technologies known in the world of web development such as HTML, CSS, Bootstrap, and Javascript. Additionally, it is important to highlight the use of Blade; this is a powerful template engine included in the Laravel framework, which allowed the construction of system views with the advantage of reusing code that is common to different sections of the application. In addition, the development of the front-end with Blade allowed the page loading speed to increase because it was possible to divide the total content of the application into small files that in Blade are known as views, which contain dynamic information that is displayed depending

226

J. L. Amaya-Carrascal et al.

on the URL address where the user is, some images of the final result of the application can be found in Figs. 3 and 4.

Fig. 3. Information system architecture.

Regarding the functionality of each module developed, the system offers the possibility of performing a complete management of the content that the user is viewing, understanding management as the union of the following actions: viewing, creating, modifying, and deleting information. It should be noted that each of these actions is associated with an existing permission in the platform, so if the authenticated user does not have the necessary permissions to act, it will not be available to that user. Likewise, the modules offer the possibility of exporting the information belonging to them. This information is exported in XLSX and PDF formats. The action of exporting information from the system is also associated with permissions. Therefore, it is not available to all registered users if they do not have the required permissions. On the other hand, it is important to point out the use of TCP communication through sockets. In this case, the laravel web sockets and laravel Echo libraries, available for Laravel, were used. This communication is vital for the monitoring module because based on the content of the messages sent and received, it is possible to make real-time graphs of the measurements of the physiological variables of heart rate, oxygen saturation, systolic blood pressure,

Development of an Information System for the Monitoring . . .

227

Fig. 4. Information system architecture.

and diastolic blood pressure. Additionally, if any of the measurements received is not within the normal range, such measurement is understood as an alteration in the neonate’s health, therefore, an alert is issued in the system and a text message is sent informing the neonate’s caregiver of such alteration. In addition to the above, and given that the records of the measurements of the physiological variables of the neonate are taken outside the system, it was necessary to expose a series of endpoints to achieve correct communication of the web application with external systems. To achieve this, an application programming interface (API) was developed using the software architecture style called representational state transfer (REST). In this API, the information received and emitted is in JavaScript object notation (JSON) format and the communication is given with the use of the hypertext transfer protocol (HTTP).

3

Discussion

The objective of this study was the development of a web information system for the monitoring of home oxygen in neonates in the city of Manizales and Villamara. This development consisted of two main parts: software development and platform development. For software development, the agile development methodology Scrum was used together with the incremental iterative development model. Although the studies reviewed on the subject of this paper do not

228

J. L. Amaya-Carrascal et al.

mention the use of agile methodologies for software development, the adoption of the agile method for project management brings with it several advantages that help in the project construction process. Among these advantages are adaptation to changes in a project, flexibility, and constant progress in the final product [7]. Additionally, by detailing this study the use that was made of methodology, and the model used, it is contributing to the generation of knowledge in the area of scientific development and also creates a reference base for future studies. On the other hand, regarding the development of the platform, it was a web application where the same technology was used for the development of the back-end and front-end. For this writing, it is necessary to detail the technologies (indicated in the results section of this same writing) used for the development of the platform, since, being a relatively large information system that will be constantly used, it is vital to ensure that the chosen technologies have features that support the workload that the system is expected to have. With this in mind, relevant characteristics of the back-end server stand out, such as data loading and access speed, information management, and security control based on roles and permissions, in addition to the use of middleware that controls access to resources and communication with both the application front-end and external applications that use the back-end to transfer information. Regarding the front-end of the system, it is not common to find this type of information in studies conducted on the subject. However, for this study, we highlight the development of views that allow the creation of the graphical interface of the application aiming to achieve a good UI/UX relationship, which facilitates the system users to manage the information and monitor the data. Similarly, for the front-end, state-of-the-art technologies were used in the market that offers features that facilitate the development and offer great functionalities to the end user of the application. Finally, with the development of an information system in the research project “Teleassistance for home oxygen monitoring in neonates in Manizales” for the management of information related to neonatal patients and the monitoring of measurements of their physiological variables, we contribute to the creation of knowledge in the field of information systems focused on the health area. In addition, the preparation of this article, which presents the description of the technologies and methodologies used, allows future researchers in engineering to rely on this type of literature for the creation of new tools and the generation of new knowledge.

4

Conclusions

This manuscript shows the development process of a web tool that corresponds to an information system that offers the possibility of monitoring the vital signs of oxygen-dependent neonates and managing information related to these neonates. For this purpose, a series of programming languages and techniques were combined with the use of an agile methodology and a software development model that allowed the construction of the platform with the help of structured work.

Development of an Information System for the Monitoring . . .

229

Additionally, the information system aims to support health professionals in the process of monitoring the evolution of neonates who are admitted and discharged from the health institution, all this, with the use of an intuitive interface and a pleasant user experience. In addition to the above, the use of different communication protocols facilitated the exchange of data between different systems, and within the same platform. This is important, because the measurements of the monitored physiological variables are not generated within the same information system, but are issued by external tools that make use of IoT for the exchange of information. In addition, it is possible to use the developed information system as a backend for future developments such as mobile applications or independent frontends that aim to keep track of the monitored vital signs as well as the alerts generated based on them.

References 1. Ali, S.I.M., Farouk, H., Sharaf, H.: A blockchain-based models for student information systems. Egypt. Inform. J. 23(2), 187–196 (2022) 2. Atlassian. Project and issue tracking software, (n.d.). (2021) 3. Varios Autores. Laravel - the php framework for web artisans, (n.d.), 1999 4. Battistello, L., Haug, A., Suzic, N., Hvam, L.: Implementation of product information management systems: identifying the challenges of the scoping phase. Comput. Ind. 133, 103533 (2021) 5. Bernal-Acevedo, O. and Forero-Camacho, J.C.: Information systems in health sector in Colombia. Revista Gerencia y Pol´ıticas de Salud 10(21), 85–100 (2011) 6. Bolin, E.H., Lang, S.M., Tang, X., Collins, R.T.: Propranolol versus digoxin in the neonate for supraventricular tachycardia (from the pediatric health information system). Am. J. Cardiol. 119(10), 1605–1610 (2017) 7. Ciric, D., Lalic, B., Gracanin, D., Tasic, N., Delic, M., Medic, N.: Agile vs. traditional approach in project management: strategies, challenges and reasons to introduce agile. Procedia Manuf. 39, 1407–1414 (2019) 8. Haried, P., Claybaugh, C., Dai, H.: Evaluation of health information systems research in information systems research: a meta-analysis. Health Inform. J. 25(1), 186–202 (2019) 9. Haux, R.: Health information systems-past, present, future. Int. J. Med. Inform. 75(3–4), 268–281 (2006) 10. Iivari, J., Hirschheim, R.: Analyzing information systems development: a comparison and analysis of eight is development approaches. Inf. Syst. 21(7), 551–575 (1996) 11. Jiang, X., Ding, Y., Ma, X., Li, X.: Compliance analysis of business information system under classified protection 2.0 of cybersecurity. Procedia Comput. Sci. 183, 87–93 (2021) 12. Nugraha, A., Daniel, D.R., Utama, A.A.G.S.: Improving multi-sport event ticketing accounting information system design through implementing RFID and blockchain technologies within COVID-19 health protocols. Heliyon 7(10), e08167 (2021)

230

J. L. Amaya-Carrascal et al.

13. O’Byrne, M.L., Glatz, A.C., Hanna, B.D., Shinohara, R.T., Gillespie, M.J., Dori, Y., Rome, J.J. and Kawut, S.M.: Predictors of catastrophic adverse outcomes in children with pulmonary hypertension undergoing cardiac catheterization: a multiinstitutional analysis from the pediatric health information systems database. J. Am. Coll. Cardiol. 66(11), 1261–1269 (2015) 14. OBS Business School. The 5 stages in scrum development sprints — obs business school, (n.d.)., 2021 15. Schwaber, K.: Scrum development process. In: Business Object Design and Implementation: OOPSLA’95 Workshop Proceedings 16 October 1995, Austin, Texas, pp. 117–134. Springer (1997) 16. Vidgen, R.: A multiple perspective approach to information system quality. University of Salford (United Kingdom) (1996) 17. Westmark, V.R.: A definition for information system survivability. In: 37th Annual Hawaii International Conference on System Sciences, 2004. Proceedings of the, 10pp. IEEE (2004)

Principal Component Analysis for Knowledge Transfer in the Social Structure Reconstruction Program in Post-conflict Zones in Colombia (Chocó, Sucre and Caldas) Marcelo López(B)

, Germán Gómez , and Carlos Marulanda

Universidad de Caldas, Calle 65 No. 26-10, Manizales, Caldas, Colombia {mlopez,germgolon,carlose}@ucaldas.edu.co

Abstract. The purpose of this writing is to present the findings of the principal component analysis of knowledge transfer from the social fabric reconstruction research program in post-conflict areas in Colombia. The study was based on the application of the principal component tool, which was founded on an instrument that allowed the evaluation of 68 research professors working with communities in the social, institutional, educational, entrepreneurial, and ecosystem dimensions that support the program. The research was based on a qualitative approach, with descriptive and correlational perspectives. The main conclusion is that knowledge transfer is taking place within the categories of Strategy, Relationship, Research, Entrepreneurship, Digital Access, and Results Publication. It is expected that these findings will be useful in establishing continuous improvement strategies within the framework of knowledge transfer from any investigative process to communities with a focus on social interventions and understanding how to design and develop effective and sustainable knowledge transfer models and processes in rural and post-conflict contexts. Keywords: Knowledge transfer · Principal components · Territorial peace

1 Introduction For several decades, Colombian universities have been facing a great number of challenges, as a result of the educational needs that have been generated, as well as the political, economic and social dynamics of the region. This is how they have had to adapt to changes related to national and international competition, the increase in coverage, the decrease in resources resulting from enrollment and government contributions, as well as the demands of a society aware of its participation in the solution of the various problems of the communities and the social challenges brought about by the peace process with irregular armed groups, under the traditional mission of teaching, research and extension enriched by generating entrepreneurship and innovation for local development. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 231–243, 2023. https://doi.org/10.1007/978-3-031-36957-5_20

232

M. López et al.

Thus, in 2017, the program reconstruction of the social fabric in post-conflict areas in Colombia was implemented, which responds to the country’s challenges in (i) social innovation for economic development and productive inclusion; (ii) quality education from science; (iii) building a stable and lasting peace, from territories and rural communities hard hit by the conflict in Chocó, Sucre and Caldas and from four lines of work: An ecosystemic model of rural improvement, entrepreneurship, social transformation and education, in addition to a transversal project of institutional strengthening, through the co-construction of research, development and innovation strategies, R&D+i, multidisciplinary and intersectoral for the strengthening of active citizen political capacities, productive competencies, capacities for peacebuilding, media literacy and generation of sustainable solutions, [1]. Within this framework, a group of professors from different public universities, among them, the Universidad de Caldas and the Universidad Nacional, Manizales, have been supporting the processes related to the program for the reconstruction of the social fabric in post-conflict areas in Colombia and have sought an important exercise related to the management and transfer of knowledge, from the professors, among the professors, the universities, the program management and the communities that have a direct relationship with the research program. And this in relation to what [2] argues, who explains that in knowledge transfer there are agents, there are actions and there are processes. Knowledge transfer must necessarily be understood as a matter of actions and not only as a matter of events and as [3] explain, Knowledge transfer and its creation, together with the experience within the framework of a learning organization, constitute today the core of those factors considered critical for the success of organizations.

2 Theoretical Framework Knowledge transfer refers to the process by which knowledge is shared, disseminated and used in different contexts and for different purposes. This process can occur between individuals, organizations and/or communities, and can involve both explicit knowledge transfer (codified in the form of documents, manuals, reports, etc.) and tacit knowledge (based on experiences, skills, interpersonal relationships, etc.). One of the most influential theoretical frameworks in knowledge transfer is the triple helix model, proposed by Ref. [4]. This model describes the interaction between three key actors: government, industry and academia. According to this model, knowledge transfer occurs through collaboration and cooperation among these three actors, and can be enhanced by specific policies and strategies. Another relevant theoretical framework is the knowledge absorption model proposed by Ref. [5], which focuses on how firms acquire and use external knowledge to improve their innovation capabilities. According to this model, a firm’s ability to absorb external knowledge depends on its capacity to recognize, assimilate, transform and exploit such knowledge. In addition, the contingency theory proposed by Ref. [6] suggests that knowledge transfer can be influenced by contextual factors, such as organizational culture, leadership and organizational structure. Likewise, Ref. [7] diffusion of innovations theory indicates that knowledge transfer can be influenced by social factors, such as risk perception, complexity and relative

Principal Component Analysis for Knowledge Transfer

233

advantage of the innovation. In the field of education, the transfer of learning theory of Ref. [8] emphasizes the importance of knowledge transfer in problem solving and decision making in different contexts. This theory suggests that transfer of learning depends on the similarity between learning contexts and application contexts, as well as on the individual’s ability to recognize and apply learned concepts and skills in new and different situations. Finally, Ref. [9] knowledge management theory describes how organizations can create, share and use knowledge to improve their performance and innovation capacity. Knowledge transfer, which, even given the time of the pandemic, needs to rely on technological tools for adequate results, as [10] state, knowledge transfer through virtual environments is a more elaborate process than face-to-face, it requires the participation of more actors and more creative resources, which is possible thanks to the existence of a series of structures and mechanisms that create, contain and transfer knowledge. Therefore, universities are called upon to direct teaching and research to the solution of social, economic, political and any other problems within their area of influence, through mechanisms that allow the transfer of knowledge, [11]. Knowledge transfer should be understood as a success factor for organizations, therefore, it is important to analyze the interrelationships of the factors that influence knowledge transfer between different disciplines to explore solutions that reduce or even avoid knowledge deficits, [12]. In such a way, knowledge transfer represents a fundamental mechanism for the achievement of any organizational activity, since it combines the knowledge of those who make it possible to achieve certain objectives and goals, from the commitment of all [13]. And it is important not only to identify, share and apply valuable knowledge within an organization, but also to consider the implications of knowledge obtained from outside it, therefore, it is essential to determine the influential factors that arise from the internal transfer of knowledge, [14]. According to Shao and Ariss [15] explain that the process of knowledge transfer involves two aspects: first, that researchers share their knowledge and, second, that their colleagues apply it. But knowledge can move through different channels (not all of which generate economic impact), such as publications, graduating students entering the labor market, academic relationships with industry and industry, and formal licensing of intellectual property to third parties, such as start-ups. Many of the products and technological advances that we take for granted in our daily lives come from university research before being transferred to the market through knowledge transfer processes, [16]. The directionality and performance (efficiency or effectiveness) of transferred knowledge affect the relationship between integration mechanisms and knowledge transfer. Directionality is especially important of integration mechanisms and knowledge transfer, [17, 18]. Some strategies for effective knowledge transfer, explain [19], can be: Dissemination of research results on radio, television, newspapers or social networks, copyright applications, patent applications, contributions to the development of clinical practices, process improvements or procedure manuals, subcontracting or creation of transfer offices. As well as: intensive communication, social networks, mobility, daily routines and communication technologies [20, 21]. In this line, [22], identified four mechanisms for knowledge transfer: (1) contribution of knowledge to organizations: (2) knowledge sharing through interactions within or between work teams; (3) informal knowledge sharing

234

M. López et al.

between individuals; and (4) knowledge sharing within communities of practice, in a much more general framework and in accordance with the role played by university teachers in their schools or faculties. Or those identified by [23], such as training, communication, observation, technology transfer, patents, partnerships, Internet databases and inter-city peer-to-peer exchange programs, and external providers and channels of communication between practitioners and researchers several decades, Colombian universities have been facing a great number of challenges, as a result of the educational needs that have been generated, as well as the political. 2.1 Knowledge Transfer and Peace Processes Knowledge transfer has become a crucial issue in peace processes, where the implementation of sustainable and lasting solutions implies access to and dissemination of relevant information. In this context, knowledge transfer can be defined as the process of bringing knowledge, skills and technologies to the people and organizations that need them to improve their capacity to act in specific situations. Knowledge transfer in peace processes has been the subject of study by several authors who have explored the role it plays in building more just and peaceful societies. Knowledge transfer is an important means of overcoming post-conflict challenges and promoting economic and social development. In addition, some authors highlight the importance of knowledge transfer in addressing social and economic exclusion, discrimination and violence in post-conflict societies [24, 25]. Knowledge transfer is a crucial process for development and social welfare, especially in contexts of armed conflict and peace processes. Knowledge transfer can help build trust, foster reconciliation and promote sustainable development. Peace processes are a set of measures and actions that seek to end armed conflict and establish a framework for peaceful and just coexistence. Knowledge transfer is essential for the success of peace processes. Knowledge transfer can help build trust between conflict actors, improve understanding of the causes and consequences of conflict, promote reconciliation, and strengthen the capacity of parties to engage in peacebuilding. Knowledge transfer can be especially important for local communities that have been affected by armed conflict and have a key role to play in peace-building. There are several factors that can influence knowledge transfer in peace processes, such as culture, politics, economics and technology. According to the above and taking into account the different mechanisms, dimensions and structures found in the literature, the interest of the researchers is to establish how the transfer process is carried out from the research program to the intervened communities, considering the most important or most appropriate variables for the reality of the universities and the communities, from the dynamics of the theoretical construct developed in the research project that supports the findings presented in this article and that is explained in all its expression in the methodology chapter.

Principal Component Analysis for Knowledge Transfer

235

3 Methodology It is based on qualitative research, with descriptive, correlational and explanatory types of study, given the multivariate analysis carried out. The theoretical construct that underlies the presentation of the results required the analysis of normality using the KolmogorovSmirnov test for small samples, followed by the spearman correlation analysis and the cronbach’s alpha analysis, fundamental tests to measure the reliability of the measurement scale. In addition to the above, a survey was constructed, which was evaluated by experts and after the pilot test, the electronic form was designed and sent to the e-mails of 90 researchers of the program “Reconstruction of the Social Fabric in Post-conflict Areas in Colombia and researchers close to the program, of which 68 responded, that is, 61% of the total population. Likewise, the instrument contains questions with answers on a Likert scale, which were evaluated from 1 to 5, where 1, do not agree, 3, agree or apply and 5, totally agree or totally apply. The variables evaluated are explained in (Table 1). Table 1. Variables Variables

Description

Direcc

The project’s strategic direction includes the knowledge transfer dimension

Saberes

The project transfers knowledge to the community based on autochthonous and promising knowledge for its transformation

Aprexpl

Project participants are encouraged to learn and explore new ways of interacting with the community and with other projects in the program

Ident

Project participants identify community needs for economic empowerment and social and productive inclusion

Relac

The relationship with the community and between projects facilitates the transfer of knowledge

Medios

There are channels, media, platforms and digital environments that support the transfer of knowledge

Compe

Project participants have the skills and capacities to transfer knowledge to communities

Gobern

A management and governance framework is in place to transfer knowledge to the community

SGKMi

The projects have a management system based on knowledge and innovation

HerrKMi

The projects incorporate methods, methodologies, techniques or tools based on knowledge management and innovation for knowledge transfer

TXcomu

Knowledge transfer is made about, from and for the community, recognizing its strengths

MarcoPI

There is a regulatory framework for both open and proprietary intellectual property (continued)

236

M. López et al. Table 1. (continued)

Variables

Description

Invcol

Collaborative, structured, inter-organizational research is or has been done

IniSNCT

There are or have been innovative initiatives and results within the framework of the technological maturity levels of the national science and technology system

Alianz

Strategic alliances are in place for innovation, entrepreneurship and knowledge transfer

Aseso

Consultancies aimed at innovation and knowledge transfer have been generated or are being developed

Consult

Have been generated in the project or are developing consultancies in knowledge management and innovation services

TransOtra

Innovation and knowledge transfer initiatives have been generated or are being developed with other organizations

PI

Intellectual property has been generated or is being developed

Empren

Spin-off ventures or Startup alliances have been generated or are being developed

Proto

Prototypes have been generated or new products or services have been developed with the community

Libro

Research books or primers or textbooks related to the activities carried out by the project have been published

Tesis

Master’s or doctoral research theses related to the project’s activities have been published

Digit

The results of research and innovation have contributed to the generation of digital content

Event

The research results have been published in specialized events such as seminars, congresses, conferences, etc

4 Results From the results found after the application of the work instrument, a general analysis of the total responses can be made, as can be seen in (Fig. 1). Most of the answers to the variables are in the framework of totally agree and only the variables SGKM+i, IniSNCT, Alianz, Consult, PI, Empren and Proto, have a qualification of not agreeing in percentages of 25% approximately, and this makes it necessary to propose alternatives that allow to improve in: projects with a management system based on knowledge and innovation, innovative initiatives and results, strategic alliances for innovation, entrepreneurship and knowledge transfer, consultancies in knowledge and innovation management services, intellectual property, Spin-off ventures or Startup alliances and prototypes or developed new products or services with the community. Now, in this framework it is possible to analyze variable by variable, but it is necessary to consider the possibility of grouping them into categories, for which it is possible to use the statistical technique of principal component analysis, which according to [26], is a widely used method for dimension reduction and the extraction of patterns from

Principal Component Analysis for Knowledge Transfer

237

Direcc Event80 Saberes Digit Apr-expl Tesis Ident 60 Libro Relac 40 Proto

Medios

20

Empren

Compe

0

PI

Gobern

TransOtra Consult Aseso Alianz IniSNCT I do not agree Fully agree or fully apply

SGKM+i HerrKM+i TXcomu MarcoPI Invcol Agree or apply

Fig. 1. Overall results

time series data. It has been applied in various fields, such as geophysics, economics, meteorology and medicine and as many naturally occurring phenomena in various fields exhibit periodic pattern changes when observed over a sufficiently long period of time, stationarity, which is a vital assumption of this analysis. In (Table 2) the results of the total variance explained can be observed. These show that the first six components account for 77.6% of the total results, i.e., the variables can be grouped into six categories that can represent the transfer of knowledge in the research program. Table 2. Total variance Component

Initial eigenvalues

Extraction sums of squared loads

Total

% Variance

Accumulated %

Total

% de Variance

Accumulated %

1

11,014

44,056

44,056

11,014

44,056

44,056

2

2,434

9,734

53,79

2,434

9,734

53,79

3

1,862

7,446

61,237

1,862

7,446

61,237

4

1,633

6,531

67,768

1,633

6,531

67,768

5

1,411

5,645

73,413

1,411

5,645

73,413

6

1,049

4,197

77,61

1,049

4,197

77,61

7

0,99

3,959

81,569

8

0,815

3,261

84,83

9

0,796

3,186

88,015

10

0,727

2,906

90,921

For a better location of the grouping of the variables, the dispersion and the associations or groupings in their principal components, the Varimax rotation technique is used, which ensures that each rotated component presents correlations between the variables.

238

M. López et al.

The rotation starts by seeking a factorial solution that is called simple structure, in terms of correlation that the value is close to 1, eliminating the negative correlations of the different variables, becoming an approximate solution (see Fig. 2).

Fig. 2. Varimax rotation

Now, the six components must be associated to the variables established in the instrument in such a way that it is fundamental to define each variable to each component, for which the structure matrix tool is used, which defines that the highest values above 0.5, are the relevant ones for each grouping, this procedure helps to minimize possible multicollinearity and allows a meaningful comparison between the variables measured with different scales. In addition, the variables that make up the interaction terms interaction terms were standardized before creating the respective cross products [27], with this analysis we can then define, according to the evaluation, the categories and variables of knowledge transfer, as shown in (Table 3). Table 3. Categories and variables 1

2

3

4

5

6

Strategy

Relationship

Research

Entrepreneurship

Digital

Results

Direcc

Aprexpl

Ident

IniSNCT

PI

Medios

Saberes

Relac

SGKMi

Empren

Digit

Tesis

Compe

Gobern

HerrKMi

Proto

TXcomu

Alianz

Invcol

MarcoPI

TransOtra

Aseso

Libro

Consult

Event

Principal Component Analysis for Knowledge Transfer

239

As explained by Ref. [28], after the exclusion of indicators with low factor loadings or that were presented in different constructs, most of the indicators are grouped together, indicating the ideal model. It means then that the transfer of knowledge in the research program can be measured from the categories: strategy, relationship, research, entrepreneurship, digital access and publication of results. Knowledge transfer strategy (component with an approximate accumulated 44%): The program and the projects resulting from its development must consider as a central strategy the transfer of knowledge to the communities, based on their knowledge and the competencies of the researchers, as well as the use of written media and taking into account intellectual property. Relationship (component with an approximate accumulated 54%): Considering the reality of a research paradigm, related to the triple helix, which is based on the participation of the State, the University and the communities to develop joint works and projects aimed at the social welfare of the regions. Research (component with an approximate 62% accumulated): From the management systems, collaborative work, tools and advisory and consulting services that universities can offer for the work of the communities, this project is an example of what can be achieved in the framework of joint work for peace in the regions of Colombia. Entrepreneurship (component with an approximate accumulated 68%): As one of the pillars of the transfer process, since the communities must consider the immense possibilities of economic development with the productive transformation of the regions, this is how linking the communities from solutions related to entrepreneurship and entrepreneurship can be of great utility for the advancement of the same. Digital access (component with an approximate accumulated 73%): In the current context, the advance of digital technology in the processes of knowledge transfer is fundamental. Publication of results (component with an approximate accumulated 78%): knowledge transfer must have results and these must be published, while opening up the possibility of continuing to work with universities in the framework of research theses, which students and professors can develop and publish in national and international events.

5 Discussions The findings presented in the previous section are supported by Ref. [29], who concludes that it is crucial to promote the transfer of scientific, technological, and innovation knowledge axis, to be worked on and strengthened between academia and the community, given that success can be achieved through sharing knowledge generated by universities. This is also affirmed by Ref. [30], who explain that multivariate analysis techniques are effective for identifying related variables through principal component analysis. Furthermore, knowledge is transferred to the university from different sources, such as personal experiences of the university team and community members. Considering that there is significant variation among different actors regarding the content of the knowledge they transfer to the university, personal experiences provide norms and the community provides information about their needs and the contexts they live in Ref. [31]. Knowledge transfer is a complex and rapidly evolving phenomenon based on the interactions of multiple stakeholders. Universities can address objectives through

240

M. López et al.

knowledge transfer activities, such as promoting innovation and practical use of research results, generating additional income streams, fostering local economic development, complying with national and institutional policies, and defining public value [32]. Regarding the work of researchers, they can argue that the benefits of knowledge transfer from research outweigh the long-term efforts for the benefit of a community that has been affected by violence for a long period and needs solutions to many difficult problems of diverse connotations. Similarly, researchers have assumed a significant additional effort related to difficulties in traveling to regions affected by violence, due to lack of infrastructure in some cases, poor road conditions, or climate problems in those regions, among others, but it is necessary to meet the schedules and objectives proposed for a successful completion of the research project [33]. It is also understood that the arduous and long processes of training and socialization of university researchers in an academic normative system generally lead to pure basic academic results, although with significant potential for contributing pragmatic solutions to a community [34]. It is from this perspective that the group of teachers has committed to learning from these communities, providing simple-to-use solutions, methods, techniques, and tools with the possibility of generating results that are necessary for advancing in the reconstruction of the social fabric, which has been deteriorated by violence. However, the analysis and results of this work pave the way for future research in this field. On an empirical level, the cross-sectional nature of the available data cannot rule out causality in the relationship between the characteristics of the university and the transfer of knowledge. In fact, although we defined the variables and categories of knowledge transfer from the university, it is necessary to consider that the characteristics of the university change in response to a particular knowledge transfer strategy. Such strategies could focus on structuring internal processes related to the protection of intellectual property, the commercialization of research, collaboration among universities, and collaboration between the university and communities. Similarly, the implementation of organizational structures and processes that adapt to knowledge transfer and the distinctive characteristics of the university should be considered.

6 Conclusions The social fabric reconstruction program in post-conflict zones in Colombia is carrying out knowledge transfer from the work of researchers, universities, and their relationship with the communities involved in the study. The knowledge transfer is being carried out within the categories of strategy, relationship building, research, entrepreneurship, digital access, and results publication. Knowledge transfer is a complex and dynamic phenomenon that is based on the interaction of various stakeholders and can be enhanced from academia to the community, and vice versa, through strategic cooperation and community dynamics. Communities can benefit from knowledge transfer, such as strengthening innovation and practical use of research results, generating additional income streams, promoting local economic development, among others. It is necessary to advance knowledge transfer from projects with a knowledge and innovation-based management system, innovative initiatives and results, strategic partnerships for innovation,

Principal Component Analysis for Knowledge Transfer

241

entrepreneurship and knowledge transfer, consultancy services in knowledge and innovation management, intellectual property, Spin-off or Startup alliances, and prototypes or new products or services developed with the community. Acknowledgments. This work is part of the results of the research program entitled: Reconstruction of the social fabric in post-conflict zones in Colombia", which was registered with the Vice-Rectory of Research and Graduate Studies of the University of Caldas and is financed with resources from the World Bank.

References 1. Cardona, L.: El programa reconstrucción del tejido social en zonas de posconflicto en Colombia participó en la primera cumbre Colombia científica. Universidad de Caldas, https://www. ucaldas.edu.co/ (2020) 2. Touriñán, J.: La transferencia de conocimiento como proceso de la universidad al sector educativo. Una mirada desde la pedagogía, Boletín Redipe 8(3), 19–652 (2019) 3. Blanco-Valbuena, C., and Pineda, W.: Transferencia de conocimiento como factor crítico para la gestión de la ciencia, la tecnología y la innovación en Maloka Bogotá, Colombia: Array. Revista Interamericana de Investigación Educación Y Pedagogía RIIEP 12(2), 41–70 (2019). https://doi.org/10.15332/25005421.5008 4. Etzkowitz, H., Leydesdorff, L.: The dynamics of innovation: from National Systems and “Mode 2” to a Triple Helix of university-industry-government relations. Res Policy 29(2), 109–123 (2000). https://doi.org/10.1016/S0048-7333(99)00055-4 5. Cohen, W.M., Levinthal, D.A.: Absorptive capacity: A new perspective on learning and innovation. Administrative. Sci. Q. 35(1), 128–152 (1990). https://doi.org/10.2307/2393553 6. Woodward, J.: Industrial organization: Theory and practice. Oxford University Press, Oxford, UK (1965) 7. Rogers, E.M: Diffusion of innovations. Simon and Schuster (2010) 8. Perkins, D.N., Salomon, G.: Are cognitive skills context-bound? Educ. Res. 18(1), 16–25 (1989) 9. Nonaka, I., Takeuchi, H.: The knowledge-creating company: How Japanese companies create the dynamics of innovation. Oxford University Press (1995) 10. Terán, A., Mendieta, B.: Modelo de transferencia de conocimiento a través de la gamificación: un gcMooc. Actualidades Investigativas en Educación 19(2) 105–131, (2019). http://dx.doi. org/https://doi.org/10.15517/aie.v19i2.36997 11. Armendáriz-Núñez, E., Tarango, J., and Machin-Mastromatteo, J.: Modelo de transferencia de conocimiento para vincular instituciones de educación superior en ciencias sociales y humanidades con entidades sociales. IE Revista De Investigación Educativa De La REDIECH (13) e1491 (2022). https://doi.org/10.33010/ie_rie_rediech.v13i0.1491 12. Klippert, M., Preißner, A., Rust, H. and Albers, A.: Analysis of factors influencing knowledge transfer between the product and production system development as well as product. Proc CIRP 109, 340–348 (2022). https://doi.org/10.1016/j.procir.2022.05.260 13. Baleta-Araujo, E.: Transferencia de conocimiento como estrategia gerencial del sector farmacéutico en Colombia. Revista Arbitrada Interdisciplinaria Koinonía, 4(8), 148–165, (2019). https://doi.org/10.35381/r.k.v4i8.275 14. Moon, H., Lee, J. H., Kim, Y., and Lee, S.: Influential factors for successful knowledge transfer in restaurant franchises. J. Hosp. Tour. Manag. 46, 224–232 (2021). https://doi.org/10.1016/ j.jhtm.2020.12.007

242

M. López et al.

15. Shao, J.J., Ariss, A.A.: Knowledge transfer between self-initiated expatriates and their organizations: Research propositions for managing SIEs. International Business Review Int. Bus. Rev. 29(1), 101634 (2020). https://doi.org/10.1016/j.ibusrev.2019.101634 16. Pohle, A., Villani, E., Grimaldi, R.: Personnel motivation in knowledge transfer offices: The role of university-level and organizational-level antecedents. Technol. Forecast. Soc. Chang. 181, 121765 (2022). https://doi.org/10.1016/j.techfore.2022.121765 17. Zeng, R., Grøgaard, B., Steel, P.: Complements or substitutes? A meta-analysis of the role of integration mechanisms for knowledge transfer in the MNE network. J. World Bus. 53(4), 415–432 (2018). https://doi.org/10.1016/j.jwb.2018.02.001 18. Díaz-Gamarra, M.: La transferencia de conocimiento en las alianzas estratégicas para la innovación. 360: Revista de Ciencias de la Gestión, (3), 136–148 (2018). https://doi.org/10. 18800/360gestion.201803.006 19. Navor-Galeana, N., Calderón-Martínez, M. G.: Nuevos indicadores de produccióny transferencia de conocimiento: una perspectiva en salud pública. Anais Do Congresso Internacional De Conhecimento E Inovação – Ciki (1), 1–16 (2020). https://doi.org/10.48090/ciki.v1i1.977 20. Wang, H., Chen, K., Zheng, H., Zhang, G., Wu, R. and Yu, X.: Knowledge transfer methods for expressing product design information and organization. J. Manuf. Syst. 58(Part A), 1–15 (2021). https://doi.org/10.1016/j.jmsy.2020.11.009 21. Majuri, M.: Inter-firm knowledge transfer in R&D project networks: A multiple case study. Technovation 15, 102475 (2022). https://doi.org/10.1016/j.technovation.2022.102475 22. Armendariz-Nuñez, E., Tarango, J., and González-Quiñones, F.: Transferencia de conocimiento en docentes universitarios y su aplicación en ciencias sociales y humanidades. Telos: revista de Estudios Interdisciplinarios en Ciencias Sociales 24 (2), 329–343 (2022). https://doi.org/10.36390/telos242.08 23. Elkhidir, E., Mannakkara, S., Henning, T. F. and Wilkinson, S.: Knowledge types and knowledge transfer mechanisms for effective resilience knowledge-sharing between cities – A case study of New Zealand. Int J Disaster Risk Reduct. 70, 102790, (2022). https://doi.org/10. 1016/j.ijdrr.2022.102790 24. Gaviria, A. and Uribe, M.: Economic growth and crime in Latin America and the Caribbean: A dynamic panel data análisis. Econ. Anal. Policy 71, 247–259, (2022). https://doi.org/10. 1016/j.eap.2021.11.003 25. Tschirgi, N.: The role of civil society in peacebuilding: a critical review. J. Peacebuilding Dev. 17(1), 1–14 (2022). https://doi.org/10.1080/15423166.2022.2001802 26. Lim, Y., Kwon, J., and Oh, H.-S.: Principal component analysis in the wavelet domain. Pattern Recognit. 119, 108096 (2021). https://doi.org/10.1016/j.patcog.2021.108096 27. López-Sáez, P., Cruz-González, J., Navas-López, J. E. and del Mar Perona-Alfageme, M.: Organizational integration mechanisms and knowledge transfer effectiveness in MNCs: The moderating role of cross-national distance. J. Int. Manag. 27 (4) (2021). https://doi.org/10. 1016/j.intman.2021.100872 28. Leite, E.D., Morales, F.X., Junior, E. R. and Cháfer, L.M.: La transmisión del conocimiento formal e informal entre los trabajadores operarios del distrito industrial: el caso del distrito agroindustrial de Anápolis. Revista de Estudios Empresariales. Segunda época, 2, 6–24 (2022). https://doi.org/10.17561/ree.n2.2022.687 29. Bautista, G.: La vinculación para la transferencia de conocimiento, científico, tecnológico e innovación, aplicado al sector mueblero de Jalisco. Paradigma económico. Revista de economía regional y sectorial, 13(2), 79–101 (2021). https://www.redalyc.org/articulo.oa? id=431567919004 30. Gallego, L., and Araque, O.: Variables de Influencia en la Capacidad de Aprendizaje. Un Análisis por Conglomerados y Componentes Principales. Información tecnológica, 30(2), 257–264 (2019). https://doi.org/10.4067/S0718-07642019000200257

Principal Component Analysis for Knowledge Transfer

243

31. Zhou, R., and Tang, P.: The role of university Knowledge Transfer Offices: Not just commercialize research outputs! Technovation 90–91, 102100 (2020). https://doi.org/10.1016/j. technovation.2019.102100 32. Giuri, P., Munari, F., Scandura, A. and Toschi, L.: The strategic orientation of universities in knowledge transfer activities. Technological Forecasting and Social Change, 138, 261–278 (2019). https://doi.org/10.1016/j.techfore.2018.09.030 33. Cartaxo, B., Pinto, G., and Soares, S.: Towards a model to transfer knowledge from software engineering research to practice. Inf. Softw. Technol. 97, 80–82 (2018). https://doi.org/10. 1016/j.infsof.2018.01.001 34. Meng, D., Li, X., and Rong, K.: Industry-to-university knowledge transfer in ecosystem-based academic entrepreneurship: Case study of automotive dynamics & control group in Tsinghua University. Technol. Forecast. Soc. Chang. 141, 249–262 (2019). https://doi.org/10.1016/j. techfore.2018.10.005

Epileptic Seizure Prediction Methods Using Machine Learning and Deep Learning Models Maria Alejandra Pati˜ no-Claros1 , Sergio Alejandro Holguin-Garc´ıa1(B) , Alvaro Eduardo Daza-Chica1 , Reinel Tabares-Soto1,2 , and Mario Alejandro Bravo-Ortiz1 1

Department of Electronics and Automation, Universidad Aut´ onoma de Manizales, Manizales (170001), Colombia {sergioa.holguin,mario.bravoo}@autonoma.edu.co 2 Universidd de Caldas, Departamento de Sistemas e Inform´ atica, Manizales (170001), Colombia

Abstract. Epilepsy is a brain disease that affects about 50 million people worldwide. It is characterized by excessive discharges in brain cells, leading the patient to have seizures, loss of consciousness, and alteration of the senses, among others. It is estimated that 5 million cases of this pathology are diagnosed annually. The most commonly used method to diagnose the disease is an electroencephalogram (EEG), which contains information about brain functions and is inexpensive. After the EEG is performed, it is observed by a professional to confirm or rule out the disease; however, this can be a delayed process, thus affecting a possible early treatment of the pathology. Those are why deep learning models have been used to create neural networks that can automatically classify this disease, thus facilitating the work of physicians and reducing time. This paper demonstrated that the deep learning model is more efficient than the machine learning model for pathology classification with an accuracy of 0.9767.

Keywords: Epilepsy Hyperparameter

1

· Deep Learning · Machine Learning ·

Introduction

Epilepsy is a brain disease that affects about 50 million people worldwide. It is characterized by excessive discharges in brain cells [13], leading the patient to have seizures, loss of consciousness, and alteration of the senses, among others. It is estimated that 5 million cases of this pathology are diagnosed annually [4]. In addition, recent studies have shown a relationship between psychiatric comorbidities and epileptic seizures, demonstrating that about 54% of patients suffer from psychological illnesses, such as depression or anxiety [12]. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 244–253, 2023. https://doi.org/10.1007/978-3-031-36957-5_21

Epileptic Seizure Prediction Methods Using Machine Learning

245

The most commonly used method to diagnose the disease is an electroencephalogram (EEG), which contains information about brain functions and is inexpensive [20]. EEG has multiple categories depending on its recording; among them is: Prolonged EEG, used to obtain sleep patterns and can last between 1 and 2 h; Ambulatory EEG is used to monitor a patient’s complete impulses throughout a typical day; this is between 1 and 3 d. However, routine EEG lasts between 20 and 30 min; as mentioned above, it is inexpensive [5]. After the EEG is performed, it is observed by a professional to confirm or rule out the disease; however, this can be a delayed process, thus affecting a possible early treatment of the pathology [15]. On the other hand, artificial intelligence has taken a crucial role in this new century, especially in biomedical engineering. Machine learning is one of the essential disciplines in this area, achieves through algorithms to identify patterns and make predictions, and in turn, as a sub-branch of this deep learning, has received much attention in this field thanks to its ability to eliminate noisecreating a boom in neuroscience and the creation of neural networks [16]. Those are why deep learning models have been used to create neural networks that can automatically classify this disease, thus facilitating the work of physicians and reducing time [21]. Using a single-channel EEG, Nipun Dilesh Perera et al. [14] developed a seizure derivation system trained with a super vector machine (SVM). They obtained a sensitivity between 87.1% and 90.16 % with a specificity close to 100% with a sample of 12 patients. In the same way, using SVM, Chun Chen et al. [7] obtained an accuracy of 97.62%. Abdulkadir Sengur et al. [17], in addition to SVM, used a feature extraction technique with an accuracy of 90.3%. But the most notable breakthrough was thanks to Arunkumar et al. [3] where they used the Bern Barcelona database and, utilizing five characteristics, achieved an accuracy of 98%, a sensitivity of 100% and a specificity of 96 %. As a deep learning method, Thara, D. K. et al. [19] achieved a specificity of 91.47% and an accuracy of 97.21% with a sensitivity of 98.59% using the PyEEG database. In this paper, we compare deep learning and machine learning models for bimodal EEG classification n epilepsy disease, predicting according to patterns whether the patient suffers from the pathology or not. The materials and methods Sect. 2 is divided into: 2.1 we will observe the description of the dataset and the modifications were made for the evaluation. In Sect. 2.2 the models used for classification are described. Section 2.3 shows and describes the metrics used. In Sect. 3 we compare the results of the methodologies used. Finally, in Sect. 4 we give the conclusions of the article.

2 2.1

Material and Methods Dataset

The epileptic seizure recognition database [2] is a modification of one of the most common databases for recognizing this pathology. It comes from an original with five different folders, each with 100 files, each representing a person and recording

246

M. A. Pati˜ no-Claros et al.

brain activity in 23.6 s. During the recording, there are 4097 data points, each representing EEG values at different time instants. There are 500 individuals and 4097 data points of 23.5 s. This database is modified, dividing and mixing each data point of 4097 in 23 chunks; each chunk contains 178 data points per 1 s. Thus obtaining 11500 pieces of information, each has 178 data points per second, and the last column represents the class (a,b,c,d,e). There are five different classes, the first class (a) is the recording of epileptic activity, the second class (b) is the EEG recording of the area where the tumor was located, the third class (c) is the EEG recording of the healthy part of the brain after locating the tumor, the fourth class (d) is the EEG recording when the individual has his eyes closed, the fifth and last class (e) is the EEG recording when the individual is with his eyes open. Considering that the only class where the epileptic activity occurred was in the first class (a), the rest of the classes (b,c,d,e) did not register epileptic activity. For this article, the database was modified; instead of having the different five classes (a,b,c,d,e) and performing a multinomial classification, it was decided to perform a binomial classification, thus retaining the class “a” where the epileptic seizure occurred but being renamed into class one (1), with a total of 2300 data points corresponding to 20% of the data, as shown in Fig. 1. With the different classes where the epileptic seizure was not present (b,c,d,e), they were grouped into one, leaving as a result, class zero (0), with a total of 9200 data points, corresponding to 80% of the database as shown in Fig. 1.

Fig. 1. Two-class database layout diagram

2.2

Model

For this project, different methods were explored, looking for the best possible alternative for the prediction of epileptic seizures using machine learning and also

Epileptic Seizure Prediction Methods Using Machine Learning

247

designing deep learning algorithms using TensorFlow software, which allows the construction of artificial neural networks (ANN). Among the different methods and algorithms used are Extra Trees Classifier (ETC), Support Vector Machine (SVM), Random Forest Classifier (RFC), Artificial Neural Network (ANN), and Gradient Boosting (GB). Extra Trees Classifier and Random Forest Classifier (ETC and RFC) These models, which are ensemble decision trees, linear classification, or regression machine learning models, tend to overfit the training data [9]. Overcome this limitation, RF trains multiple decision trees. Each decision tree uses a random subset of the sample and a subset of features from the entire training set to achieve greater tree diversity, leading to better results [6]. Extra Trees Classifier models, on the other hand, add randomness to the model training process by using random decision thresholds for each feature instead of searching for the best possible threshold. Support Vector Machine (SVM) Solve linear or nonlinear classification and regression. This model is best suited for small and medium complex datasets. The basic idea of SVM classification is to separate classes that respect decision boundaries as much as possible from the closest training pattern. Keep as far as possible from the closest training pattern Add nonlinearity. SVMs [9] create linearly separated classes using kernel functions that modify or add features depending on the training set [10]. Gradient Boosting (GB) Boosting is any set method combining multiple weak learners into a single strong learner [8]. The general idea is to train predictors in order, each trying to correct the previous predictor. GB is a popular boosting algorithm that attempts to fit new predictors to the residuals of old predictors [9]. Artificial Neural Network (ANN) also called multilayer perceptron, defines a description y = f (x; θ) (1) and learns parameter values or weights, theta that achieve the best approximation of the inputs “x” to the output “y” [11]. The ANN layer consists of several units that compute a linear operation on the input. Then, triggered by a nonlinear function such as “ReLU”, “SeLU” or “TanH,” it is activated, and that output is sent sequentially to the next layer until the data reaches the output layer. The backtracking calculates the prediction error, and the network parameters are updated to reduce it. The backtracking process followed by a backward step is repeated for a fixed number of iterations or until convergence [1]. This model is shown in Fig 2.

248

M. A. Pati˜ no-Claros et al.

Input features

Conv1D

Conv1D

Conv1D

Filters: 32 Kernel size: 1 Strides: 1 Padding: Valid Activation: ReLU

Filters: 32 Kernel size: 1 Strides: 1 Padding: Valid Activation: ReLU

Filters: 32 Kernel size: 1 Strides: 1 Padding: Valid Activation: ReLU

Dense Neurons: 512 Activation: SELU

Dense Neurons: 256 Activation: SELU

Dense Neurons: 128 Activation: SELU

Dense Neurons: 64 Activation: SELU

Flatten

Dense Neurons: 6 Activation: SELU

Class probabilities Flow patterns

Softmax

Fig. 2. Artificial Neural Network Architecture (ANN)

Fig. 3. Hyperparameters of the different neural networks classified in deep learning and machine learning

2.3

Metrics

Tabares-Soto et al. [18] explain the importance of metrics for evaluating a model and differentiate between false positives (FP), false negatives (FN), true positives (TP), and true negatives (TN). The metrics used for this Evaluation are shower in Fig. 3. The most important metrics are the following: Accuracy Accuracy is the fraction ranging from 0 to 1, representing the correct prediction percentage. Accuracy =

TP + TN TP + TN + FP + FN

(2)

Epileptic Seizure Prediction Methods Using Machine Learning

249

Precision Metrics are used to identify the correct distribution of classes. TP (3) TP + FP Recall Also known as sensitivity, it shows the ability of the classifier to display correct predictions. P recision =

TP (4) TP + FN F1 Parameter is used to favor class imbalance, as it is a weighting between accuracy and recall. Recall =

P resicion ∗ Recall (5) P resicion + Recall Support This metric indicates the number of data in each test class. Confusion Matrix The confusion matrix is the combination of the actual and predicted classes. The rows represent the envisioned classes, and the columns represent the real class. F1 = 2 ∗

3

Results

Table 1 shows the total accuracy report of each of the models used, both machine learning and deep learning, and the precision of these models in classifying the different classes that were defined, being “Epileptic seizure” class one and “Without Epileptic seizure” class zero. In class one, a better accuracy result was obtained with the Gradient Boosting model, with a value of 0.973. A higher precision result is evidenced in the deep learning model with the artificial neural network (ANN) being 0.9761; with this same model, a better precision result was also obtained with a value of 0.984 in the classification of class zero. Table 1. Total accuracy report and precision of both classes Accuracy [Ac] SVM RF GBC ETC ANN

0.8000 0.9739 0.9704 0.9757 0.9761

Epileptic Without Epileptic SD Precision 0.3 0.1 0.5 0.3 0.4

0.000 0.969 0.973 0.965 0.964

Precision 0.800 0.975 0.970 0.978 0.984

In the other models used, accuracy and precision values very similar to those already mentioned were obtained, except for the classification performed by Support Vector Machine (SVM) in which the lowest values were received, with precision in class one of 0.000 and class zero a value of 0.800 was obtained, the same value obtained in the accuracy of this model.

250

M. A. Pati˜ no-Claros et al.

Figure 4 shows the recall, f1, and support accuracy metrics for classes one and zero (a) and confusion matrices (b) of the Gradient Boosting (GB) machine learning models and deep learning using Artificial Neural Network (ANN).

(a)

(b)

Fig. 4. Metrics of Gradient Boosting Classifier of machine learning model (a), Confusion Matrix of Gradient Boosting Classifier (b)

Figure 5 shows the learning curve of the artificial neural network, for which 5iteration cross-validation was performed in which the 11500 data set was divided into 0.8 for training (9200) and the remaining 0.2 (2300) for testing. In this curve, it is evident success obtained in the classification with the artificial neural network model.

4

Conclusions

This research work represents the different results for the classification of epilepsy disease, showing the best alternatives for this specific problem. The database was modified to accurately separate the classes that do not present epileptic activity from the class that does present epileptic activity.

Epileptic Seizure Prediction Methods Using Machine Learning

251

Fig. 5. Learning curve of ANN

As part of the design of the classification system, some key aspects were identified, such as database, hyperparameter settings, deep learning architecture construction (ANN), and metrics evaluation. Despite the remarkable ability of machine learning models to process and classify data, hyperparameter fitting ensures better performance. The selection of hyperparameter fitting models is an iterative process that allows finding the optimal configuration for each model in a given evaluation range. The main contribution of this work is to compare machine and deep learning models to find the best alternatives. For deep learning, it is necessary to find a balance between computational capacity, throughput time, and model performance. An extensive neural network should perform better, but in practice, more extensive networks increase the training time, thus increasing the computational load, with tendencies to overtrain. With this in mind, it was decided to use an artificial neural network (ANN), a small network designed to optimize the performance and efficiency of the model, which was trained to 70 epochs, as shown in the learning curve (Fig. ??). It obtained an overall accuracy of 0.9761 and a precision of 0.964 for class one and 0.984 for class zero. On the other hand, in the use of machine learning, the least efficient of the methods used is the Support Vector Machine (SVM), a machine learning model that failed to classify the first class with a value of 0.000 and both the overall accuracy and the accuracy of the zero class obtained a value of 0.800. The most efficient machine learning method was Gradient Boosting (GBC), with a total accuracy of 0.9704.

252

M. A. Pati˜ no-Claros et al.

As future work, it is proposed to initially perform a data balancing to subsequently train the algorithm using the five classes of the original database in order to achieve better accuracy in the prediction of epileptic seizures.

References 1. Abiodun, O.I., Jantan, A., Omolara, A.E., Dada, K.V., Mohamed, N.A., Arshad, H.: State-of-the-art in artificial neural network applications: a survey. Heliyon 4(11), e00938 (2018) 2. Andrzejak, R.G., Lehnertz, K., Mormann, F., Rieke, C., David, P., Elger, C.E.: Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: dependence on recording region and brain state. Phys. Rev. E 64(6), 061907 (2001) 3. Arunkumar, N., Ramkumar, K., Venkatraman, V., Abdulhay, E., Fernandes, S.L., Kadry, S., Segal, S.: Classification of focal and non focal EEG using entropies. Pattern Recognit. Lett. 94, 112–117 (2017) 4. Beghi, E.: The epidemiology of epilepsy. Neuroepidemiology 54(2), 185–191 (2020) 5. Benbadis, S.R., Beniczky, S., Bertram, E., MacIver, S., Mosh´e, S.L.: The role of EEG in patients with suspected epilepsy. Epileptic Disord. 22(2), 143–155 (2020) 6. Breiman, L.: Random forests. Mach. Learn. 45, 5–32 (2001) 7. Chen, C., Liu, Z., Li, H., Zhou, R., Zhang, Y., Liu, R.: EEG detection based on wavelet transform and SVM method. In: 2016 IEEE International Conference on Smart Cloud (SmartCloud), pp. 241–247. IEEE (2016) 8. Friedman, J.H.: Stochastic gradient boosting. Comput. Stat. Data Anal. 38(4), 367–378 (2002) 9. G´eron, A.: Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow. O’Reilly Media, Inc. (2022) 10. Gholami, R., Fakhari, N.: Support vector machine: principles, parameters, and applications. In: Handbook of Neural Computation, pp. 515–535. Elsevier (2017) 11. Gu, J., Wang, Z., Kuen, J., Ma, L., Shahroudy, A., Shuai, B., Liu, T., Wang, X., Wang, G., Cai, J., et al.: Recent advances in convolutional neural networks. Pattern Recognit. 77, 354–377 (2018) 12. Lu, E., Pyatka, N., Burant, C.J., Sajatovic, M.: Systematic literature review of psychiatric comorbidities in adults with epilepsy. J. Clin. Neurol. (Seoul, Korea) 17(2), 176 (2021) 13. Organization, W.H., et al.: Epilepsy: A Public Health Imperative. World Health Organization (2019) 14. Perera, N.D., Madarasingha, C., De Silva, A.C.: Spatial feature reduction in longterm EEG for patient-specific epileptic seizure event detection. In: Proceedings of the 9th International Conference on Signal Processing Systems, pp. 230–234 (2017) 15. San-Segundo, R., Gil-Mart´ın, M., D’Haro-Enr´ıquez, L.F., Pardo, J.M.: Classification of epileptic EEG recordings using signal transforms and convolutional neural networks. Comput. Biol. Med. 109, 148–158 (2019) 16. Saxe, A., Nelli, S., Summerfield, C.: If deep learning is the answer, what is the question? Nat. Rev. Neurosci. 22(1), 55–67 (2021) 17. S ¸ eng¨ ur, A., Guo, Y., Akbulut, Y.: Time-frequency texture descriptors of EEG signals for efficient detection of epileptic seizure. Brain Inform. 3, 101–108 (2016)

Epileptic Seizure Prediction Methods Using Machine Learning

253

18. Tabares-Soto, R., Arteaga-Arteaga, H.B., Mora-Rubio, A., Bravo-Ort´ız, M.A., Arias-Garz´ on, D., Alzate-Grisales, J.A., Orozco-Arias, S., Isaza, G., Ramos-Poll´ an, R.: Sensitivity of deep learning applied to spatial image steganalysis. PeerJ Comput. Sci. 7, e616 (2021) 19. Thara, D., PremaSudha, B., Xiong, F.: Auto-detection of epileptic seizure events using deep neural network with different feature scaling techniques. Pattern Recognit. Lett. 128, 544–550 (2019) ¨ Ozerdem, ¨ 20. T¨ urk, O., M.S.: Epilepsy detection by using scalogram based convolutional neural network from EEG signals. Brain Sci. 9(5), 115 (2019) 21. Xu, G., Ren, T., Chen, Y., Che, W.: A one-dimensional cnn-lstm model for epileptic seizure recognition using EEG signal analysis. Front. Neurosci. 14, 578126 (2020)

Towards Energy Efficiency in Microgrids for Smart Sustainable Cities V. Isanbaev1 , R. Ba˜ nos1(B) , C. Gil2 , M. M. Gil1 , F. Mart´ınez1 , and A. Alcayde1 1

2

Department Engineering, University of Almer´ıa, Carretera de Sacramento s/n, La Ca˜ nada de San Urbano, 04120 Alme´ıa, Spain {vs613,rbanos,fmg714,alcayde}@ual.es Department Informatics, University of Almer´ıa, Carretera de Sacramento s/n, La Ca˜ nada de San Urbano, 04120 Alme´ıa, Spain [email protected]

Abstract. Microgrids are a critical component of smart and sustainable cities as they provide localized power generation and distribution that can be optimized for efficiency, cost, and environmental impact. Further, many microgrids are characterized for the presence of smart homes, which are an integral part of smart cities as they play a crucial role in improving the overall efficiency and sustainability of urban areas. One of the primary benefits of smart homes is the ability to reduce energy consumption and carbon emissions, for example, by automatically adjusting the temperature, lighting, and other settings to optimize energy usage based on the users’ needs and preferences. However, the efficient management of smart homes located in microgrids is still an open question. In particular, the efficient management of microgrids including smart homes requires measuring and processing a large amount of electrical data related to the energy generated by the power sources of the microgrid, the energy consumed by the loads (home appliances), the level of battery storage and the amount of transferred power flow between the microgrid and the main grid. This paper reflects on how to measure electrical variables at different points of the microgrids using low-cost smart meters with IoT capabilities and how to apply Non-Intrusive Load Monitoring (NILM) methods for energy efficiency purposes. The empirical study involving a real microgrid including a hybrid wind-solar generation system and a set of home appliances show that it is possible to collect data of different parts of the microgrid for efficient management purposes.

Keywords: Microgrids Smart Meters · IoT



· Smart Sustainable Cities · Smart homes ·

This Research Was Funded by the Ministry of Science, Innovation and Universities, Grant Number PGC2018-098813-B-C33 and ERDF Funds

c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 254–265, 2023. https://doi.org/10.1007/978-3-031-36957-5_22

Towards Energy Efficiency in Microgrids for Smart

1

255

Introduction

Smart cities, microgrids, and smart homes are all interrelated concepts in the realm of sustainable urban development. A smart city integrates technology and data-driven solutions to enhance the quality of life for its citizens and improve urban sustainability. One way a smart city can achieve this is by implementing microgrids, which are self-contained energy systems that can operate independently from the main power grid. Smart homes are residential buildings that use advanced technology and automation to optimize energy usage and reduce waste. By integrating smart homes in microgrids, it is possible to create a more sustainable, efficient, and resilient urban environment. Further, smart homes can use energy from a microgrid during peak demand periods and feed excess energy back into the grid when demand is low. This paper presents a framework to monitor and analyze energy and power quality parameters in microgrids integrating home devices using smart meters. The rest of the paper is organized as follows. Section 2 presents some key projections on the future of smart and sustainable cities and microgrids, and reflects on the importance of designing advanced electricity metering procedures in these contexts. Section 3 presents a framework to monitor and analyze energy consumption of home appliances in order to apply NILM methods for energy efficiency purposes. Section 4 presents a case study in a real environment where it is shown how this framework can be implemented. Finally, Sect. 5 presents the conclusions and future investigations derived from this study.

2

Related Work

The future of smart and sustainable cities [1] is exciting, with many predictions and expectations for how they will evolve and develop over the coming years. Some of the key predictions for smart cities include the widespread adoption of smart technology, the integration of renewable energy sources, and the development of intelligent transportation systems. Other predictions include the development of more sophisticated data analysis tools [2] and the continued growth of the Internet of Things (IoT), with more devices and infrastructure being connected to the Internet. More attention is also likely to be given to optimising energy use and promoting energy efficiency to reduce carbon emissions. Some ways in which smart cities promote energy-saving [3] are: – Smart grids: Smart cities use smart grids to optimize the distribution of energy. Smart grids use sensors and data analysis to monitor energy usage and adjust the supply of energy accordingly. This can help reduce energy waste and promote energy efficiency. – Energy management systems: Smart cities use energy management systems to monitor energy consumption and optimize energy usage in buildings and infrastructure. These systems can adjust heating, cooling, and lighting based on occupancy levels and other factors, reducing energy waste and promoting energy efficiency.

256

V. Isanbaev et al.

– Renewable energy: Smart cities promote the use of renewable energy sources such as solar and wind power. By generating renewable energy locally, smart cities can reduce reliance on fossil fuels and promote energy-saving. – Smart transportation: Smart cities use smart transportation systems to promoting public transportation, electric vehicles, and intelligent traffic management systems to reduce congestion, which wastes fuel and entails a big loss of time for citizens [4]. – Data Analysis: Smart cities use data analysis to identify areas of high energy consumption and optimize energy usage. By collecting and analyzing data on energy usage, smart cities can identify opportunities for energy-saving and promote energy efficiency. A microgrid [5] is a small-scale electrical grid that can operate independently or in conjunction with the larger power grid using local generation sources such as solar panels, wind turbines, and energy storage systems. One of the key benefits of microgrids in smart cities is related to the energy reliability and resiliency. By generating and distributing electricity at the local level, microgrids can continue to provide power to critical infrastructure and services during grid outages or natural disasters. Another advantage of microgrids is their ability to integrate renewable energy sources into the energy mix. By using solar panels, wind turbines, and other renewable sources, microgrids can reduce greenhouse gas emissions and improve air quality in urban areas, which is particularly important because cities are major sources of greenhouse gas emissions. In addition to these benefits, microgrids can also help to reduce energy costs by optimizing energy usage and reducing peak demand on the larger power grid. This can lead to cost savings for both consumers and utilities, as well as reducing the strain on the larger power grid during peak usage periods. Other important element of sustainable smart cities are smart homes [6], that is, homes equipped with smart devices and appliances that can communicate with each other and the grid, providing users with more control over their energy usage and costs. Microgrids can incorporate smart homes in several ways [7], enabling them to optimize energy usage and increase energy efficiency. One way microgrids can incorporate smart homes is by allowing homeowners to generate their own electricity using renewable energy sources and to use it to power the home appliances, with any excess electricity being stored in batteries or fed back into the grid. This allows homeowners to reduce their reliance on the larger power grid and potentially save money on their energy bills. Microgrids can also incorporate smart homes by enabling homeowners to monitor and control their energy usage in real-time. This can be done through smart meters and energy management systems that provide homeowners with data on their energy usage and costs. By analyzing this data, homeowners can identify areas where the energy usage can be reduced to save money. In addition, microgrids can allow smart homes to participate in demand response programs, where homeowners can reduce their energy usage during periods of high demand on the grid. This helps to reduce strain on the grid and potentially earn homeowners incentives or credits for participating. Finally, microgrids can also incorporate

Towards Energy Efficiency in Microgrids for Smart

257

smart homes by enabling them to be more resilient since it is possible to generate and store electricity during power outages. Some investigations have analyzed energy management systems in integrated building and microgrid systems using optimization methods [8]. Figure 1 shows a typical microgrid connected to the main grid through the point of common coupling (PCC), that includes distributed generation (hybrid wind-solar generation in this case), storage system (batteries), control system, and loads (e.g. smart home appliances).

Fig. 1. Microgrid with hybrid wind and solar generation.

Measuring energy consumption of home appliances is essential for smart homes located in microgrids. Here are some key reasons why measuring energy consumption in home appliances is important: – Energy efficiency and cost savings: By monitoring the energy consumption of home appliances, smart homes can optimize energy usage and reduce waste. For example, if it is noticed that a particular appliance is consuming more energy than necessary, the homeowner can adjust the settings or replace it with a more energy-efficient model. This can help reduce energy bills and promote sustainability.

258

V. Isanbaev et al.

– Smart control: Measuring energy consumption allows smart homes to control appliances more intelligently. For example, the homeowners can use the information about consumption of appliances in combination with smart devices and automation systems to turn them off when not in use or adjust their settings remotely to reduce energy usage. – Cyber-security: Microgrid operational efficiency requires a variety of regulating techniques, sensors, actuators, as well as sophisticated but most critically secure measurement, communication, and control. This involve a large number of entry points, which is why microgrids are a highly sensitive cyberphysical system [9]. Therefore, it is necessary to implement procedures to measure the usage of the appliances in order to detect cyber-attacks. – Sustainability: Reducing energy consumption in home appliances is crucial for promoting sustainability and reducing carbon emissions. By measuring energy consumption, smart homes can help homeowners make more sustainable choices and reduce their carbon footprint. – Cost savings: Monitoring energy consumption can help homeowners to identify appliances that are consuming too much energy and replace them with more energy-efficient models. This can lead to significant cost savings over time, as energy-efficient appliances are typically less expensive to operate than older, less efficient models.

3

Monitoring Energy Data in Microgrids Using Smart Meters

Smart meters are advanced devices that can monitor and measure energy parameters of typical home appliances, making them a valuable tool for promoting energy efficiency and reducing waste. Here are the steps to monitor and measure energy parameters in typical home devices using smart meters: – Smart meter installation: The first step is to install one or several smart meters in the microgrid. The meters are located according to different strategies. In particular, Intrusive Load Monitoring (ILM) techniques [10] are based on the use of low-end electricity meter devices directly measuring each device, such that smart plugs or other devices communicate with the smart meter to send information about the energy usage in real-time. Alternatively, NonIntrusive Load Monitoring (NILM) techniques [11,12], which requires using not one meter in each device, but only a single meter which measures the overall demand across several appliances. – Monitor energy usage: Once the smart meter(s) measure(s) the energy consumed by the appliances, it is possible to monitor energy usage in real-time. Many smart meters come with apps or web portals that allow you to track energy usage and receive alerts when energy usage is high or when devices are consuming more energy than usual. – Optimize energy usage: Armed with real-time energy usage data, you can optimize energy usage by adjusting device settings, turning off devices when

Towards Energy Efficiency in Microgrids for Smart

259

not in use, or replacing old devices with newer, more energy-efficient models. This can help reduce energy waste and promote energy efficiency. In order to measure the energy parameters of the home appliances, the openZmeter device [13] is used. As Fig. 2 shows, a single openZmeter device acquires and processes energy data, which is later submited to a computer, which is in charge of processing the information received.

Fig. 2. Measuring energy parameters in home appliances using the openZmeter.

The main advantage of this measurement scheme is the use of Non-Intrusive Load Monitoring (NILM) [14] in order to measure the energy parameters of the home appliances. NILM is a technique used to disaggregate the total power consumption of a home or building into individual appliance-level power consumption data. The advantages of NILM in home devices include cost-effectiveness, ease of installation, and privacy preservation. Unlike traditional appliance-level monitoring, NILM does not require additional sensors or hardware to be installed in each individual appliance, which reduces the cost of installation and maintenance. Moreover, NILM can be easily installed in existing homes without the need for rewiring or significant modifications. Additionally, NILM preserves the privacy of the occupants by not requiring the installation of monitoring devices in each individual appliance. Furthermore, NILM can be used to identify energyintensive appliances and optimize their energy consumption, leading to energy savings and reduced electricity bills. The openZmeter device [13] has proven to work very efficiently for measuring electrical data related to home appliances [15], so thanks to its low cost, small size, and easy installation, it can be installed in different parts of the microgrid, so that power flows between the microgrid and the main grid, between distributed generation and batteries, or at the device level can be monitored.

260

4

V. Isanbaev et al.

Case Study

This section presents a case study carried out in a microgrid located at the University of Almera (Spain) (geographic coordinates: Latitude: 36◦ 49’45” N, Longitude: 2◦ 24’28” W) and described in Fig. 3.

Fig. 3. Microgrid (Aerial image and scheme of the internal building).

This microgrid consists of three wind turbines and two solar trackers, which are used to capture wind and solar energy respectively to generate electricity. The energy generated is then stored in batteries to be used during periods of low energy production or high energy demand. The batteries in the microgrid are responsible for managing and storing energy generated by the renewable

Towards Energy Efficiency in Microgrids for Smart

261

sources, ensuring a steady supply of electricity to the loads. The set of loads in the microgrid include a large number of home appliances such as refrigerators, air conditioners, and lights, among others. These loads consume energy from the batteries and the renewable sources when there is available power. The controller manages the charging and discharging of batteries, the distribution of energy to loads, and the switching between renewable energy sources depending on their availability. The advantages of a microgrid consisting of three low-power wind turbines with two solar trackers, batteries, and a set of loads including different home appliances include reduced dependence on the grid, improved energy security, and reduced carbon footprint. The microgrid can provide electricity to the smart home without access to the grid and can also be used as a backup power supply during emergencies. The openZmeter device is able to collect data about the main electrical parameters in real time. For example, Fig. 4 shows the energy (kWh), voltage (V), current (A), active power (kW) and frequency (Hz) of the entire microgrid.

Fig. 4. Main electrical variables in the microgrid measured by the openZmeter.

Since it is possible to connect many electrical devices in a microgrid, it is important to measure not only the active power but also the reactive power. Active or real power (measured in watts, W) is the electrical power that is consumed by a device and converted into useful work (e.g. producing heat, light, or motion). On the other hand, reactive power (measured in volt-amperes-reactive, VAR) is the portion of electrical power that does not perform useful work but instead is required by some devices (e.g. motors, transformers, and fluorescent lamps) to create and maintain magnetic fields. Figure 5 shows the active and reactive power of different home appliances measured by openZmeter. It is impor-

262

V. Isanbaev et al.

tant to notice that reducing the reactive power in an electrical system is important because it leads to a more efficient use of energy, resulting in lower energy costs, reduced equipment wear, and an increased capacity of the electrical grid, which can improve overall system reliability and stability.

Fig. 5. Active and reactive power of some home appliances.

Moreover, the openZmeter is also able to obtain power quality data of great importance in the microgrid. More specifically, openZmeter registers the events according the EN50160 power quality standard. For example, Fig. 6 shows the voltage swell or dips, short interruptions, long interruptions, harmonic distortion, among other parameters registered by openZmeter in the microgrid. It is also important to remark that the data retrieved by openZmeter can be later processed to obtain additional information. For example, openZmeter implements an adapted version of the Fourier transform algorithm that obtains harmonic values in complex form up to order 50, for current and voltage. Figure 7 shows the main odd harmonics with respect to the active power when openZmeter separately measures the home appliances. The openZmeter device registers any power quality event and displays the CBEMA/ITIC curve, which is a graph often used to represent voltage events and the range of voltage which typically can be tolerate by electronic devices.

5

Conclusions

Measuring energy and power quality parameters within a microgrid that powers a smart home is crucial for ensuring efficient and reliable energy supply to support smart cities. Smart meters play a critical role in collecting real-time data on energy consumption, production, and power quality parameters, such as voltage, current, power factor, harmonics, and other electrical parameters. This paper outlines how to monitor data using a set of low-cost smart meters with IoT functionalities to obtain measures from a microgrid that includes a hybrid windsolar generation system, a battery, and a set of home appliances. The results

Towards Energy Efficiency in Microgrids for Smart

263

Fig. 6. Power quality information retrieved by openZmeter according to EN50160 power quality standard.

Fig. 7. Representation of odd harmonics with respect to the active power.

264

V. Isanbaev et al.

demonstrate that data can be collected from various parts of the microgrid, including home appliances, to facilitate efficient management purposes. In particular, some conclusions about the importance of measuring energy and power quality parameters in microgrids with a smart home for smart cities using smart meters are: (1) Smart meters can provide valuable insights into how much energy is generated and consumed. This data can help homeowners and utility companies to better manage energy usage, reduce waste, and optimize energy production to meet the demand; (2) Smart meters can also monitor the power quality parameters in real-time, providing insight into the quality of the electrical supply. This information can be used to identify and address any power quality issues that could damage to appliances or lead to energy waste; (3) Measuring energy and power quality parameters can help to ensure the reliability and resilience of the microgrid, for example, by detecting faults or failures in the system and alert homeowners or utility companies, allowing them to take corrective action before a major outage occurs; (4) By monitoring energy usage, microgrid operators can identify opportunities to reduce energy consumption, optimize energy production, and reduce energy costs; and (5) Measuring energy and power quality parameters is essential for implementing sustainable energy solutions in smart cities. Smart meters can help to track energy usage, identify energy waste, and promote the use of renewable energy sources, reducing carbon emissions and promoting a cleaner environment. Some investigations have remarked that machine learning and based models are very suitable solutions for predicting consumers demands and energy generations from renewable energy sources [16]. This is why, as future work, we plan to investigate how to integrate high-dimensional data using machine learning and prediction techniques [17]. The goal is to apply different methods, including soft computing techniques [18,19], to optimize energy management by leveraging information from power sources, storage systems, and energy-consuming appliances in the microgrid. Acknowledgments. This work was supported by project PGC2018-098813-B-C33 (Spanish “Ministerio de Ciencia, Innovaci´ on y Universidades”), and by European Regional Development Funds (ERDF).

References 1. Bibri, S.E., Krogstie, J.: Smart sustainable cities of the future: an extensive interdisciplinary literature review. Sustain. Cities Soc. 31, 183–212 (2017) 2. Pinto-Santos, F., Shoeibi, N., Rivas, A., Hern´ andez, G., Chamoso, P., De La Prieta, F.: Distributed platform for the extraction and analysis of information. In: Corchado, J.M., Trabelsi, S. (eds.) Sustainable Smart Cities and Territories, pp. 200–210. Springer (2022) 3. Wang, S.J., Moriarty, P.: Energy savings from smart cities: a critical analysis. Energy Procedia 158, 3271–3276 (2019) 4. Corchado, J.M., Larriba-Pey, J.L., Chamoso-Santos, P., De la Prieta-Pintado, F.: Advances in public transport platform for the development of sustainability cities. Electronics 10(22), 2771 (2021)

Towards Energy Efficiency in Microgrids for Smart

265

5. Saeed, M.H., Fangzong, W., Kalwar, B.A., Iqbal, S.: A review on microgrids’ challenges & perspectives. IEEE Access 9, 166502–166517 (2021) 6. Kim, H., Choi, H., Kang, H., An, J., Yeom, S., Hong, T.: A systematic review of the smart energy conservation system: from smart homes to sustainable smart cities. Renew. Sustain. Energy Rev. 140, 110755 (2021) 7. Mansouri, S.A., Ahmarinejad, A., Nematbakhsh, E., Javadi, M.S., Jordehi, A.R., Catalao, J.: Energy management in microgrids including smart homes: a multiobjective approach. Sustain. Cities Soc. 69, 102852 (2021) 8. Anvari–Moghaddam, A., Guerrero, J.M., Vasquez, J.C., Monsef, H., Rahimi-Kian, A.: Efficient energy management for a grid–tied residential microgrid. IET Gener., Transm. Distrib. 11(11), 2752–2761 (2017) 9. Canaan, B., Colicchio, B., Ould-Abdeslam, D.: Microgrid cyber-security: review and challenges toward resilience. Appl. Sci. 10(16), 5649 (2020) 10. Ridi, A., Gisler, C., Hennebert, J.: A survey on intrusive load monitoring for appliance recognition. In: Proceedings of 22nd International Conference on Pattern Recognition, pp. 3702–3707. IEEE, Stockholm (Sweden) (2014) 11. Hart, G.W.: Nonintrusive appliance load monitoring. Proc. IEEE 80(12), 1870– 1891 (1992) 12. Alcal´ a, J., Ure˜ na, J., Hern´ andez, A., Gualda, D.: Event-based energy disaggregation algorithm for activity monitoring from a single-point sensor. IEEE Trans. Instrum. Meas. 66(10), 2615–2626 (2017) 13. Viciana, E., Alcayde, A., Montoya, F.G., Ba˜ nos, R., Arrabal-Campos, F.M., Manzano-Agugliaro, F.: An open hardware design for internet of things power quality and energy saving solutions. Sensors 19(3), 627 (2019) 14. Ghosh, S., Chatterjee, A., Chatterjee, D.: An improved load feature extraction technique for smart Homes using fuzzy-based NILM. IEEE Trans. Instrum. Meas. 70, 1–9 (2021) 15. Isanbaev, V., Ba˜ nos, R., Arrabal-Campos, F.M., Gil, C., Montoya, F.G., Alcayde, A.: A comparative study on pretreatment methods and dimensionality reduction techniques for energy data disaggregation in home appliances. Adv. Eng. Inform. 54, 101805 (2022) 16. Aslam, S., Herodotou, H., Mohsin, S.M., Javaid, N., Ashraf, N., Aslam, S.: A survey on deep learning methods for power load and renewable energy forecasting in smart microgrids. Renew. Sustain. Energy Rev. 144, 110992 (2021) 17. Gonz´ alez-Briones, A., Hernandez, G., Corchado, J.M., Omatu, S., Mohamad, M.S.: Machine learning models for electricity consumption forecasting: a review. In: 2019 2nd International Conference on Computer Applications & Information Security (ICCAIS), pp. 1–6. IEEE (2019) 18. Ba˜ nos, R., Gil, C., Agulleiro, J.I., Reca, J.: A memetic algorithm for water distribution network design. In: Saad et al. (eds.) Soft Computing in Industrial Applications: Recent Trends, pp. 279–289. Springer (2007) 19. Dashtdar, M., Flah, A., Hosseinimoghadam, S.M.S., El-Fergany, A.: Frequency control of the islanded microgrid including energy storage using soft computing. Sci. Rep. 12(1), 20409 (2022)

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion in Medium-Sized Latin American City Streets Yamila S. Grassi1(B)

and Mónica F. Díaz1,2

1 Planta Piloto de Ingeniería Química—PLAPIQUI (UNS-CONICET), Bahía Blanca, Argentina

[email protected] 2 Departamento de Ingeniería Química, Universidad Nacional de Sur (UNS), Bahía Blanca,

Argentina

Abstract. Air pollution sources have changed over time, but they remain a major global concern. Mobile sources, mainly on-road, are considered one of the main anthropogenic sources of several polluting gasses, particulate matter and greenhouse gasses. In this regard, population and motorized traffic growth in cities make it necessary to have tools for measuring or estimating air quality. In this sense, emission dispersion models help to estimate the pollutant concentrations in the emission sources’ surroundings, especially in cities that do not have air quality monitoring. This is the case of the Bahía Blanca (Argentina), a medium-sized Latin American city, which has high vehicular traffic in the downtown area during peak hours. Consequently, in previous works, the AERMOD model was used to estimate air quality in some downtown streets, and these results must be validated. In this work we show an initial validation proposal for the modeling methodology. Therefore, we took advantage of an air quality monitoring station recently located in an area compatible with the city center. Although the station is located on the city’s outskirts, it is an area influenced by vehicle traffic emissions, as is the city’s microcenter. The results of the present work follow the station data trend, confirming that the modeling may be able to show changes in air quality. This work is a preliminary study, and tries to find a simple and easy solution to be applied in places of interest in cities where urban air pollution levels are to be estimated in a reliable way. Keywords: Vehicular traffic · Air pollution · AERMOD · Bahía Blanca

1 Introduction Air pollution is a current concern in cities around the world [1]. Although the origin of urban air pollution has changed over time, the following aspects are distinguishable in all cases: an emission source, the pollutants themselves, a transport means—air, water, soil—and a receptor [2, 3]. At present, the population growth and the increase of motorized traffic in the cities, causes a rise in the levels of urban air pollution [1]. In this sense, urban traffic affects the urban environment and human health [4]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 266–276, 2023. https://doi.org/10.1007/978-3-031-36957-5_23

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion

267

The main instruments used to study and evaluate air quality are monitoring programs, air quality models, and emission inventories [5]. The objective of air quality modeling is to find a relationship between released substances into the atmosphere, and the concentrations of these compounds at specific receptors [6]. These models are widely used to evaluate control strategies aimed at reducing air pollution, to help in the design of effective policies, and to fulfill legislated air quality objectives [7, 8]. It is known that air quality models have the ability to describe the atmospheric dispersion of inert, reactive species and particulate matter at various scales (local, regional or continental), for which it is necessary to have information of the study area such as meteorology and geography [6]. The results obtained are of importance especially when there is no air quality monitoring network or programs designed to measure pollutant gasses in situ that allow effectively knowing the local reality [9]. This situation represents the current reality of the downtown area of Bahía Blanca, Argentina, which has no air quality monitoring. Mobile sources are considered diffuse or non-point sources because they cannot be attributed to a specific geographic location, making their monitoring and control much more difficult [2]. It is known that, combustion originated by the transport sector is one of the main anthropogenic sources of emissions of carbon monoxide (CO), nitrogen oxides (NOx) and volatile organic compounds (VOCs) in addition to sulfur dioxide (SO2 ), particulate matter (PM10 and PM2.5 ) and unburned hydrocarbons [5, 6]. Accordingly, urban air pollution levels have been reduced in cities during the COVID-19 pandemic lockdown, mainly due to the reduction of motorized vehicle traffic [10]. Consequently, cities are adopting new trends in urban mobility, considering the importance of being more people-centered and environmentally friendly [11]. In this sense, an aspect that cannot be left aside is that sustainable urban mobility is an important element in the creation of resilient and smart cities [12], which is also linked to the concept of smart environment [13, 14] and to the 2030 Agenda for Sustainable Development proposed by the United Nations [15]. In addition, it should be considered that a city is not only smart because of the technology it uses but also because of the intelligent use of procedures to make community life much easier [16]. Bahía Blanca city has experienced an unplanned urban growth, resulting in higher traffic levels in the city streets due to the increased use of private cars, mainly in the downtown area. According to preliminary data from the 2022 census, the city has a population of 335190 inhabitants [17]. In addition, it was reported for the year 2018 that the city had a vehicle fleet of approximately 172000 units [18]. At this point, it should be noted that the city’s downtown area has no monitoring of atmospheric pollutants. For this reason, in previous works [19], we have modeled the dispersion of pollutants in some microcenter city streets using the AERMOD model (United States Environmental Protection Agency—US EPA—version) to obtain estimates of pollutant gas concentration levels produced by mobile sources (urban traffic). In this regard, the estimations obtained should be validated. Recently, in the year 2022, an air quality monitoring station was placed on the outskirts of the city, at 5 km from the microcenter, in an area with moderate vehicular traffic. Therefore, we have seen this scenario as an opportunity to validate the methodology applied in previous work [19]. In this regard, the present work attempts to show how this new monitoring location and the values reported by the air quality station were used to obtain a preliminary validation of the air pollution modeling

268

Y. S. Grassi and M. F. Díaz

using AERMOD. The results of this validation will serve as a basis for the modeling carried out in areas close to and similar to this point but which do not have a monitoring station.

2 Study Area The city of Bahía Blanca is located in the south of the province of Buenos Aires, Argentina. It is a medium-sized port city that has a great interest in the study of air pollution, mainly because it has a large petrochemical complex. In this sense, Bahía Blanca city has incorporated, since the 90’s, equipment for air quality monitoring such as mobile stations equipped with air pollutant continuous analyzers (CO, NO2 , SO2 , O3 , PM10 and PM2.5 ). In addition, in 2000 the Comité Técnico Ejecutivo (CTE) was created to control and oversee the industries located in the petrochemical complex. Currently, there are two fully equipped stations and another one that monitors only particulate matter at an access point to the city. These stations (EMCABBs) were usually located near the petrochemical complex and the industrial area of the city, about 10 km from downtown, to control emissions generated by large-scale industries (point sources). However, during 2022, one of the stations (EMCABB1) was relocated to an urban area in the southern part of the city, about 5 km from the center (see Fig. 1).

Fig. 1. Current location, in Bahía Blanca city, of the air quality monitoring station (EMCABB1) used in this work.

In the new location, appropriate conditions were detected to perform a preliminary validation of the modeling that was being carried out in the city’s downtown area associated with emissions from vehicular traffic [19]. The reason for this is that the new EMCABB environment is delimited by streets with moderate traffic flow and without any relevant point sources in its surroundings. Hence, this ensures that the data recorded by the station corresponds to an air quality consistent with the downtown area and not to an industrial environment. Nevertheless, it should be noted that this site is not central

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion

269

and has different characteristics from the typical urban downtown environment, as there are no buildings in the proximity (see Fig. 2). The current location of the EMCABB is type “grassland”, which will have to be taken into account during modeling. It was not until the end of 2022 that the setting up of the monitoring equipment and the necessary calibrations were finally completed. In consequence, the present study started at the end of November 2022 when we confirmed that the station was working properly. For this reason, we have only surveyed two periods so far.

Fig. 2. Location of the air quality monitoring station (EMCABB1), where the surrounding landscape can be observed.

3 Methodology and Data Analysis Emission dispersion near a source, as mobile sources, can be modeled using Gaussian models [20]. Although a large number of freely available atmospheric dispersion models are available from the US-EPA, AERMOD is the highly recommended model [8]. For this work, version 19191 of the AERMOD model (downloaded from the official US-EPA website) has been used, together with the BETA features for the R-LINE source type. R-Line models streets as line segments with emphasis in estimating concentrations very close to the source line [21, 22]. In this sense it should be known that both AERMOD and R-LINE report better results than other models, thus together they represent a big potential [23, 24]. The methodology of this study is based on detailed fieldwork. First, videos were made in the area under study to obtain real data on the vehicular flow, as well as segmentation. The categorization was carried out taking into account 6 types of vehicles: motorcycles, cars, pickup vehicles, light commercial vehicles (LCVs), heavy trucks and buses. The videos were recorded on the morning of November 29, 2022 and February 1, 2023. This information, together with the emission factors and street dimensions, are used to determine the emission rates (ERs) as in Grassi and Díaz [19], necessary for the modeling. The emission factors (EF) were obtained from COPERT software version 5.3,

270

Y. S. Grassi and M. F. Díaz

considering the type of vehicle, the fuel used and the emission control technology (EURO standards) by vehicle age. The main AERMOD output file, in addition to containing all the established input parameters and meteorological data, includes the concentrations of the specified pollutant reaching the receptors, considering the worst-case meteorological scenario. In this work output files were generated for NOx, evaluated as a 1-h average. It should be noted that the station measures NO2 while the model calculates total NOx. For this reason, a Tier 1 approach is used, in which a complete conversion of all nitrogen oxide to nitrogen dioxide is considered [25]. For this work only the location where the monitoring station is placed was selected as the receptor (see Fig. 3).

Fig. 3. Street segmentation around the monitoring station included in the modeling presented in this paper. The length of the street under consideration is shown in parentheses.

The meteorological files needed to work with AERMOD were developed using the AERMET preprocessor (version 19191), which was also downloaded from the USEPA website. This preprocessor requires as input data the hourly surface meteorological observations and the sounding which provides information on the vertical structure of the atmosphere [26]. The meteorological data used in each case belong to the day and hour in which the videos to obtain the vehicular flow were taken. This is important because these are the specific moments in which the data generated by the monitoring station were captured from the Bahía Blanca environmental control website [27]. It should be considered that the NO2 concentration values are validated the following year by CTE technical staff and those data are not yet published. For this reason, we only have the online data from the website, and it is a screenshot at the time of the video recordings, as mentioned above.

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion

271

The vertical profile sounding data used, extracted from the US National Oceanic and Atmospheric Administration (NOAA), are those belonging to the city of Santa Rosa (La Pampa) since the Argentine National Meteorological Service (SMN) does not perform this measurement in Bahía Blanca [28]. While the integrated surface hourly database (ISHD) for Bahía Blanca is also obtained from the NOAA [29]. It should be noted that these data are measured at the city’s airport by the SMN, which is about 10 km from where the station is located. In this sense, it was also modeled using wind speed, wind direction and temperature data provided by the MeteoBahia (MB) for an area closer to the monitoring point (Pampa Central neighborhood—about 3 km from the EMCABB1 but with similar characteristics). These meteorological data are generated by the “Centro de Recursos Naturales Renovables de la Zona Semiárida (CERZOS—CONICET-UNS)” and published on their web page [30]. It should be considered that these data were compared with that generated by the station at the monitoring point, provided by CTE, to identify the source that has a better estimation of the local meteorology. Finally, it should be noted that the necessary parameters in AERMET, such as albedo, Bowen’s ratio and surface roughness, were set to the values recommended for “Grassland” in the AERMET User’s Guide [26]. This is important to note since the studied site does not have the same characteristics as the microcenter of the city (see Fig. 2). In this sense, the albedo, Bowen ratio and surface roughness data were set at: 0.18, 0.80 and 0.10 for summer (December to February); 0.20, 1.00 and 0.01 for autumn (March to May); 0.20, 1.50 and 0.01 for winter (June to August); and 0.18, 0.40 and 0.05 for spring (October to December).

4 Results and Discussions This section presents the results achieved on vehicular flow and segmentation as well as modeling and meteorology. It should be noted that the results presented are limited, i.e., they are representative of two specific days and hours. Consequently, this study should be considered preliminary, and as a basis to continue validating in each season of the year and in other time periods. This work arises as an opportunity at the end of the year 2022, when a continuous air quality monitoring station is located in an area of Bahía Blanca city, 5 km from the city center, that is influenced by emissions generated by mobile sources (urban motorized traffic), similar to the center. In this sense, the findings of validating the method using the data collected from the station will allow the model to be used in other similar city areas that do not have air quality monitoring. 4.1 Vehicular Flow of the Study Area and Its Segmentation This subsection analyzes the vehicular flow and segmentation of the study area. Table 1 presents the collected data of vehicles per hour on both monitored days, considering the segmentation of the streets as presented in Fig. 3. In general, it can be observed that in February all streets have experienced a decrease in vehicular traffic, which was to be expected since it is one of the months of the summer holidays. The largest decreases are detected in the car and pickup truck categories with an overall average of −20% in February, reaching −40% in some streets. In addition, in the heavy truck segment, the

272

Y. S. Grassi and M. F. Díaz

highest overall variation was −127% in February compared to November. This situation can also be observed in the global segmentation by type of vehicle carried out over the study area in Fig. 4, where heavy trucks represented a larger portion of the total number of vehicles in November than in February. On the other hand, cars, motorcycles and pickup trucks are the dominant vehicles in the study area, a situation similar to that found in the microcenter of Bahía Blanca city as reported in previous studies [11, 19, 31]. Table 1. Categorization of vehicle flow, in vehicles per hour, that passed through the study area on the days and times analyzed, considering the street segmentation presented in Fig. 3. PP1

PP2

AA1

AA2

AA3*

AA4*

DM

68

12

40

98

98

80

2022/11/29–8 a.m. to 9 a.m Motorcycles

44

Cars

310

370

92

290

322

322

316

Pickup trucks

56

68

28

66

74

74

58

LCVs

66

40

2

40

42

42

12

Buses

8

6

4

4

6

6

0

Heavy trucks

18

14

4

22

16

16

4

2023/02/01–8 a.m. to 9 a.m Motorcycles

46

50

32

38

74

74

44

Cars

266

220

66

226

304

304

188

Pickup trucks

54

52

16

50

60

60

44

LCVs

50

30

4

36

38

38

8

Buses

8

4

6

8

8

8

0

Heavy trucks

6

4

2

0

0

0

0

* The street segments labeled AA3 and AA4 have the same vehicular flows; they are separated only to simplify the area modeling.

Fig. 4. Global categorization of the vehicle fleet that circulated in the study area on the dates considered in this study.

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion

273

4.2 Meteorology and Modeling in the Study Area As previously mentioned, the modeling has been performed considering two sources of meteorological data for the wind and temperature variables. Table 2 shows them and allows us to see the differences that exist, mainly in the wind variable. These changes have an impact on the model since wind speed is one of the main factors to be considered in the transport of pollutant gasses in the atmosphere [6]. It is interesting to note that the wind speed and direction differ among the three sources, but not so much the temperature. In this sense, comparing the three sources, it is observed that the meteorological conditions reported by MeteoBahia are more similar to those registered by EMCABB1 than the ones recorded by the Argentine National Meteorological Service. Table 2. Wind speed and temperature data for the analyzed dates, considering the three information sources: National Meteorological Service (SMN), MeteoBahia (MB) and the air quality station (EMCABB1). Date

Meteorology

Wind speed [km/h (m/s)]

Wind direction

Temperature [°C]

2022/11/29 8:00 a.m

SMN

13.0 (3.6)

NE

21.5

2022/11/29 9:00 a.m

MB

5.7 (1.6)

NNE

23.7

EMCABB1

5.9 (1.6)

ONO

24.1

SMN

7.0 (1.9)

NO

22.7

MB EMCABB1

2023/02/01 8:00 a.m

2023/02/01 9:00 a.m

SMN

15.2 (4.2)

NO

26.4

9.9 (2.7)

O

27.4

15.0 (4.1)

O

18.0

MB

11.4 (3.1)

ENE

19.6

EMCABB1

11.7 (3.2)

E

S/D

SMN

20.0 (5.6)

O

19.9

MB EMCABB1

9.6 (2.6)

ESE

20.9

12.9 (3.6)

ENE

S/D

As for the results obtained from the modeling, it can be said that they are uneven and do not allow a direct conclusion (see Table 3). It was expected to have a better fit when applying the data from MeteoBahia, however, only in the case analyzed in November a value relatively close to that recorded by the station was obtained. At this point it is necessary to remember that this work is preliminary, since only two very specific moments have been analyzed, for which the vehicle flow data and its segmentation were obtained. Although the results are preliminary, it should be noted that the modeling follows the trend of the data generated by the air quality monitoring station. As the concentrations recorded by the station decreased in February, the modeling followed this trend. It should be noted that this assessment can be made considering the Tier 1 approach mentioned in the methodology in which all NOx is converted into NO2 .

274

Y. S. Grassi and M. F. Díaz

Table 3. Results obtained from the modeling on the dates analyzed and comparison with the data recorded by the air quality monitoring station. Date

Meteorology

Station NO2 concentration (1 h) [µg/m3 (ppb)]

Modeled NOx concentration (1 h) [µg/m3 (ppb)]

Difference (%)

2022/11/29

SMN

20.13 (10.69)

13.49 (7.17)

−33

MB

20.13 (10.69)

19.15 (10.17)

−5

2023/02/01

SMN

5.95 (3.16)

0.52 (0.27)

−91

MB

5.95 (3.16)

1.15 (0.61)

−80

As mentioned above, it is observed that the model can estimate trends in nitrogen oxides concentrations near mobile sources. Likewise, it is evident that the variation of only two meteorological variables affects the modeling, as does the variation of vehicular flow and its segmentation. In this sense, it should be noted that a model obtains an estimate of reality and in this case not only the wind speed and temperature affect the modeling but also all the meteorological variables as well as the parameters set related to them (the albedo, Bowen ratio and surface roughness). On the other hand, the emission factors have certain limitations as they are estimates of the emissions of each type of vehicle considered, and these are used on the emission rates. In addition, we must not forget that the data validated by the CTE has not yet been published and we are awaiting this confirmation to continue our work. In this sense, there are many factors involved in the modeling of pollutant gas dispersion, for this reason we consider these preliminary results as promising, since they follow the trend of what was recorded by the in-situ air quality monitoring station.

5 Conclusions As has been previously mentioned, it is important to have information on urban air quality. In this sense, those cities that do not have continuous or in situ monitoring can employ modeling to estimate the concentration values of several polluting gases, but it is necessary to validate the results. This work attempts to find a simple and easy solution to validate pollutant dispersion modeling using data from an air quality monitoring station located at a site affected by urban motorized traffic, with a similar scenario than those without air quality measures. It should be noted that the AERMOD model has previously been employed in the city center of Bahía Blanca, but the estimations obtained could not be validated because there were no air quality measurements in that location. In this sense, this work emerges as an opportunity to validate the methodology and applicability of the AERMOD model in Bahía Blanca City, considering a new monitoring point 5 km from the city center but influenced by motorized traffic, similar to center, and where a continuous air quality monitoring station was recently installed. Although this work is an initial validation, the first modeling results suggest that the trend of the monitoring data is being followed, which looks promising and encourages the continuation of this research.

Initial Validation Regarding the Use of AERMOD to Model Air Pollutant Dispersion

275

Acknowledgments. We are grateful to Mr. Leandro Lucci (B.Sc. in Biochemistry), member of the Comité Técnico Ejecutivo (CTE) of the Municipality of Bahía Blanca, for providing essential data for this work. Also, we are particularly thankful to the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), for the funding granted to this research through the PIP project [Grant N° 11220210100683CO]. Additionally, this work was partially supported by the Universidad Nacional del Sur (UNS), Bahía Blanca, Argentina, [Grant N° PGI 24/M158].

References 1. Gulia, S., Nagendra, S., Khare, M., Khanna, I.: Urban air quality management—A review. Atmos. Pollut. Res. 6(2), 286–304 (2015). https://doi.org/10.5094/APR.2015.033 2. Alloway, B., Ayres, D.: Chemical principles of environmental pollution, 1st edn. Blackie Academic & Professional, Glasgow (1993) 3. Thornbush, M.: Vehicular air pollution and urban sustainability: an assessment from Central Oxford, UK. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-20657-8 4. Tang, V., Oanh, N., Rene, E., Binh, T.: Analysis of roadside air pollutant concentrations and potential health risk of exposure in Hanoi, Vietnam. J. Environ. Sci. Health, Part A 55(8), 975–988 (2020). https://doi.org/10.1080/10934529.2020.1763091 5. Gallego Picó, A., et al.: Contaminación atmosférica. UNED, Madrid (2012) 6. Porta, A., Sánchez, E., Colman Lerner, J.: Calidad de Aire. Monitoreo y modelado de contaminantes atmosféricos. Efectos en la Salud Pública. Editorial de la Universidad de La Plata, La Plata (2018) 7. de la Guardia, M., Armenta, S.: The Quality of Air. Comprehensive Analytical Chemistry— Volume 73. 1st edn. Elsevier, Amsterdam (2016) 8. US-EPA, Air Quality Models. https://www.epa.gov/scram/air-quality-models, last accessed 2023/02/07 9. Querol, X.: La calidad del aire en las ciudades: Un reto mundial. Fundación Gas Natural Fenora, Madrid (2018) 10. Sokhi, R., et al.: A global observational analysis to understand changes in air quality during exceptionally low anthropogenic emission conditions. Environ. Int. 157, 106818 (2021). https://doi.org/10.1016/j.envint.2021.106818 11. Grassi, Y., Díaz, M.: Post-pandemic urban mobility in a medium-sized Latin American city. Is sustainable micro-mobility gaining ground? Int. J. Environ. Stud. (2023). https://doi.org/ 10.1080/00207233.2023.2195327 12. Ceder, A.: Urban mobility and public transport: future perspectives and review. Int. J. Urban Sci. 25(4), 455–479 (2021). https://doi.org/10.1080/12265934.2020.1799846 13. Barbarossa, L.: The post pandemic city: Challenges and opportunities for a non-motorized urban environment, An overview of Italian cases. Sustainability 12(17), 7172 (2020). https:// doi.org/10.3390/su12177172 14. Alderete, M.: Determinants of smart city commitment among citizens from a middle city in Argentina. Smart Cities. 4, 1113–1129 (2021). https://doi.org/10.3390/smartcities4030059 15. United Nations, Sustainable Development Goals. https://www.un.org/sustainabledevelop ment/, last accessed 2023/02/08 16. Rom, J.: Charla TEDx: La smart city de los ciudadanos inteligentes. https://www.youtube. com/watch?v=_3NzZm2bygs, last accessed 2022/06/23 17. INDEC, Censo 2022. https://www.censo.gob.ar/, last accessed 2023/02/06 18. Grassi, Y., Brignole, N., Díaz, M.: Vehicular fleet characterisation and assessment of the on-road mobile source emission inventory of a Latin American intermediate city. Sci. Total Environ. 792, 148255 (2021). https://doi.org/10.1016/j.scitotenv.2021.148255

276

Y. S. Grassi and M. F. Díaz

19. Grassi, Y., Brignole, N., Díaz, M.: Pandemic impact on air pollution and mobility in a Latin American medium-size city. Int. J. Environ. Stud. 79(4), 624–650 (2022). https://doi.org/10. 1080/00207233.2021.1941662 20. Forehead, H., Huynh, N.: Review of modelling air pollution from traffic at street-level—The state of the science. Environ. Pollut. 241, 775–786 (2018). https://doi.org/10.1016/j.envpol. 2018.06.019 21. Snyder, M., Venkatram, A., Heist, D., Perry, S., Petersen, W., Isakov, V.: RLINE: A line source dispersion model for near-surface releases. Atmos. Environ. 77, 748–756 (2013). https://doi. org/10.1016/j.atmosenv.2013.05.074 22. Valencia, A., Venkatram, A., Heist, D., Carruthers, D., Arunachalam, S.: Development and evaluation of the R-LINE model algorithms to account for chemical transformation in the near-road environment. Transp. Res. Part D: Transp. Environ. 59, 464–477 (2018). https:// doi.org/10.1016/j.trd.2018.01.028 23. Heist, D., et al.: Estimating near-road pollutant dispersion: A model inter-comparison. Transp. Res. Part D: Transp. Environ. 25, 93–105 (2013). https://doi.org/10.1016/j.trd.2013.09.003 24. Stocker, J., Heist, D., Hood, C., Isakov, V., Carruthers, D., Perry, S., Snyder, M., Venkatram, A., Arunachalam, S.: Road source model intercomparison study using new and existing datasets. In 15º Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes (HARMO), Madrid (2013). https://www.harmo.org/Conferences/ Proceedings/_Madrid/publishedSections/H15-78.pdf, last accessed 2023/02/06 25. US-EPA, Nitrogen Dioxide/Nitrogen Oxide In-Stack Ratio (ISR) Database. https://www.epa. gov/scram/nitrogen-dioxidenitrogen-oxide-stack-ratio-isr-database, last accessed 2023/02/01 26. US-EPA. User’s Guide for the AERMOD Meteorological Preprocessor (AERMET). EPA454/B-21-004, April 2021 27. Control Ambiental Bahía Blanca (Comité Técnico Ejecutivo – CTE). https://cte.controlam biental.bahia.gob.ar/calidad-de-aire/. last accessed 2023/02/01 28. NOAA, ESRL Radiosonde Database – upper air sounding. https://ruc.noaa.gov/raobs/, last accessed 2023/02/13 29. NOAA, Integrated Surface Hourly Database. https://www1.ncdc.noaa.gov/pub/data/noaa/, last accessed 2023/02/13 30. MeteoBahia. https://www.meteobahia.com.ar/index.php, last accessed 2023/02/01 31. Grassi, Y., Brignole N., Díaz, M.: Análisis del tránsito vehicular, pre y post pandemia, en el microcentro de la ciudad de Bahía Blanca (Argentina) orientado hacia la sustentabilidad. In Congreso Internacional Universidad, Territorio y Ambiente. Universidad Nacional de Quilmes, Bernal (2022). https://www.youtube.com/watch?v=4YPR24Kh74Y

Evaluation of Feature Descriptors for Scene Classification Luis Hernando R´ıos Gonz´alez1,3 , Sebasti´ an L´ opez Fl´ orez1,2(B) , 1,2 Alfonso Gonz´ alez-Briones , and Fernando de la Prieta1,2 1

3

BISITE Digital Innovation Hub, University of Salamanca, Edificio Multiusos I+D+i, Calle Espejo 2, 37007 Salamanca, Spain {lhgonza,sebastianlopezflorez,alfonsogb,fer}@usal.es 2 Air Institute, IoT Digital Innovation Hub, 37188 Salamanca, Spain Universidad Tecnol´ ogica de Pereira Cra. 27 N 10-02, Pereira, Risaralda, Colombia

Abstract. The current article discusses the performance of local and global descriptors, as well as convolutional neural networks (CNNs), in tasks involving image recognition in interior spaces. The purpose of the test is to identify several realistic situations that closely resemble the typical working conditions for mobile robots. A robot interacting with its environment may be able to see portions of scenes in which objects are seen from various angles or changes in the lighting in various settings. The purpose is to investigate how well the different descriptors perform in identifying situations that meet the above criteria. In order to evaluate the effectiveness of visual descriptors and convolutional neural networks in the classification of images taken from the perspective of mobile robots in indoor environments, a proprietary database was implemented and subjected to several controlled transformations. These modifications made it possible to analyze the performance of Bag-of-Visual-Words (BoVW), Fisher Vectors (Fisher), Vector of Locally Aggregated Descriptors (VLAD), Global Image Descriptors (GIST), and CNN descriptors in visual categorization tasks according to the situational perception of mobile robots.The findings highlight the advantages of descriptors for the various test scenarios and highlight the need for hybrid models that employ both descriptors and CNNs for scene identification tasks in interior areas where mobile robots operate. Keywords: Visual descriptors · Scene categorization Deep learning · Gist recognition

1

· Robotics ·

Introduction

The interaction between humans and automated systems is increasingly empathetic as a consequence of the great advances in information technology, communication and the Internet of things, as well as the development of computer c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 277–288, 2023. https://doi.org/10.1007/978-3-031-36957-5_24

278

L. Hernando R´ıos Gonz´ alez et al.

systems.An autonomous mobile robot must analyze each action, the robot’s current state, and the next action that has to be conducted. One of an autonomous mobile robot’s perceptual abilities is vision, where the significance of recognizing a scene can offer a straightforward explanation of an image’s content. However, because this process considers not only the existence of a global context but also the semantic relationships between objects and contextual information, it may be challenging for a robot to perceive its immediate surroundings [24]. The first descriptors were inspired by research on human perception of the scene. Their goal was to capture the dominant spatial structure and other properties on which humans focus to classify scenes. Typical global attribute descriptors are Global Image Descriptor (GIST) [14] and its variants. For these, global features capture the diagnostic structure of the image, providing an approximate version of the main contours and textures of the image that is still detailed enough to recognize the essence of the scene. The extraction of feature patches for tasks such as object recognition has extensively used local visual descriptors such as patch-based coding descriptors, such as Scale Invariant Feature Transform(SIFT) [12], Speeded-Up Robust Features(SURF) [3], Histogram of oriented gradients (HOG) [6]. These descriptors are useful for capturing textures, shapes and other distinctive spatial structures and are used to encode pattern information in image patches. Gradient histograms, which are invariant to scale and rotation, are used to describe image regions in the neighborhood of selected keypoints. The SIFT descriptor encodes orientation and magnitude gradients in the neighborhood of each keypoint. Unlike global descriptors that provide a holistic representation of a scene, patch-based descriptors capture local information [20]. These descriptors can be conveniently interrelated to model the objects of interest present in a given image. When local image descriptors such as SIFT, SURF or HOG are used, it is necessary to aggregate them in a fixed dimensionality compact form to generate an adequate (feature vector) representation of each image that can then be fed into a statistical classifier to recognize the scene [21]. Such aggregation is not necessary when global image descriptors such as Pyramid Histogram of Oriented Gradients (PHOG) or GIST are used instead. The Bag of Visual Words (BoVW) [21] method provides a compact, invariant representation of images, capable of capturing subtle differences in content while maintaining semantics. By encoding images as histograms of ”visual words” rather than pixels, it produces more general representations that are robust to variations in illumination, scale, viewpoint, etc [10]. Once the bag of visual words representing the distribution of visual words in an image is created, it encodes the semantic signature used to train a statistical classifier and learn to discriminate between different categories of scenes. Different classifiers can be used, such as Support Vector Machines (SVM) or Naive Bayes Classifiers [5]. Finally, the results of recognition consist of the scene category, the scene name or additional related information. Promising results have been achieved with BoVW systems [4,5] for object category recognition. However, one of the main disadvantages of the BoVW model is that generating the dictionary does not take into account the geometric relationships between visual words.

Evaluation of Feature Descriptors for Scene Classification

279

Techniques based on learned codebook reconstruction offer a promising approach to overcome the limitations of conventional aggregation methods for learning bag-of-words feature representations. By modeling the structure of the input space and reconstructing it from the learned features or codes, these techniques can capture more semantic structure and be more robust to outliers and distortions [19]. Where hard quantization, in which each local visual word is assigned to a single signature. However, it can fail when local visual descriptors resemble several signatures. To solve this problem, soft assignment methods such as the Gaussian mixture model (GMM) [26] are used. The GMM associates descriptors with multiple similar words proportionally. Recently, the Fisher kernel [17] and the Vector of Locally Aggregated Descriptors (VLAD) [9] have been explored as alternative soft coding methods. These soft alternatives to hard quantization can capture the subtleties and similarities of descriptors related to several words. However, they also tend to be more computationally intensive due to the use of probability distributions or residual aggregations instead of direct assignments. Scene recognition models based on convolutional neural networks(CNNs) [16, 18] have achieved promising results, outperforming the bag of visual words and other traditional approaches on some datasets and benchmark tasks. The hierarchical architecture of CNNs allows them to automatically learn powerful global functions suitable for scene recognition. The evolution of CNN architectures has followed a trend towards deeper and more powerful models, from the relatively shallow LeNet-5, proposed in 1998, to AlexNet [16], deeper with 8 layers, proposed in 2012, which achieved significant improvements in image classification and catalyzed modern CNN research. ZFNet [25] and VGGNet [25] were proposed in 2014, based on AlexNet with greater depth (13–19 layers) and other modifications. GoogleNet, proposed in 2014, introduced an “inception module” that increased performance through a novel structure. It won ImageNet 2014. ResNet [8], proposed in 2015, is very deep (up to 152 layers) with “residual connections” that helped training. It won ImageNet 2015. Increasing depth allows the network to better approximate the objective function with greater nonlinearity and obtain better feature representations. However, it also increases network complexity, making the network harder to optimize and more prone to overfitting. This article evaluates methods for recognizing static indoor scene images in the face of robotics challenges. Derived from an experimental and methodological analysis of visual descriptors for scene categorization. The analysis of visual descriptors includes both local and global perspectives.

2

Feature Representations

The main framework for scene recognition starts by extrapolating the characteristics from the entrance picture. The transformation of characteristics is then used to capture some of the key elements of the scene and create a representation of the image. Finally, classification is done by a trained classifier, for instance, vector support machines, into semantic categories. Therefore, the

280

L. Hernando R´ıos Gonz´ alez et al.

approach of recognizing an image is based on the conventional method of extrapolating local characteristics using an image representation and then classifying them.The trend towards deep features and end-to-end deep learning is due to the ability of DCNNs to automatically learn hierarchical feature representations directly from data. These learned features have been shown to achieve better results than hand-designed alternatives in a number of computer vision tasks. Some end-to-end deep learning networks unify feature extraction, feature transformation, and image classification into a coherent pipeline. 2.1

Descriptors of Global Characteristics

GIST These descriptors allow identifying visual attributes such as texture and geometric pattern. The GIST model encodes the spatial structure of a scene using a set of oriented Gabor filters. A Gabor filter is a Gaussian envelope modulated by a cosine function, which produces a bandpass filter. Gabor filters are applied at multiple scales and orientations, capturing patterns at different levels of detail and different orientations. Then, the responses of Gabor filters at different scales and orientations are grouped into a two-dimensional representation of the scene [21]. This global representation encodes the dominant spatial properties of the scene, such as openness, roughness and naturalness. It does not capture precise object categories or fine details, but shows that coarse spatial structure alone can be informative for distinguishing scene categories. GIST features are also interpretable and related to perceptual properties, unlike the complex learned features of deep networks. The impulse response of a Gabor filter is a Gaussian modulated by a harmonic function: x G(x, y) = cos(2π + φ) γ





x 2 + γ2y 2 exp( ) 2σ 2

(1)

where x and y are the coordinates of the spatial domain x = xcosθ + ysinθ, y  = −xsinθ + xcosθ, σ where θ is the rotation angle, determines the scale of the Gaussian envelope, φ is the orientation of the normal to the parallel bands of the Gabor function, γ is the wavelength of the harmonic function. The responses of Gabor filters at all scales and orientations are combined to produce a low-dimensional representation of the spatial structure of the scene. Classification is performed using a linear support vector machine (SVM). The GIST approach shows that spatial structure alone can be informative for scene categorization. However, GIST features are limited in capturing object recognition and fine-grained details. Their performance has been surpassed by deep convolutional neural networks, but they remain interpretable and useful for understanding scene structure [14]. Feature encoding using Deep Networks Deep learning architectures are neural networks with multiple hidden layers between input and output layers. These layers transform input data into increasingly abstract representations through successive applications of nonlinear functions. The power of deep

Evaluation of Feature Descriptors for Scene Classification

281

learning comes from its ability to automatically learn hierarchical feature representations directly from data. Deep learning architectures are neural networks with multiple hidden layers between input and output layers. These layers transform input data into increasingly abstract representations through successive applications of nonlinear functions. The power of deep learning comes from its ability to automatically learn hierarchical feature representations directly from data, relieving the need for hand-designed features. CNNs have achieved state-of-the-art results in computer vision, primarily presented by LeCun et al. [11]. A convolutional neural network (CNN) is a feedforward architecture with three main types of layers: – Convolutional layers: which applies filters to capture spatial patterns in the inputs, the results of each filter are feature maps indicating the presence of features in different locations. – Pooling layers: reduce the size of each feature map by taking the maximum or average of map regions. This handles distortions and reduces the dimensionality of features. – Fully connected layers: traditional multilayer perceptrons that use the pooled features from convolutional and pooling layers to produce outputs CNNs have been used to recognize static indoor images representing complex environments in which robots must make decisions based on place recognition. The authors in [22] apply transfer learning with a convolutional neural network (CNN) to the task of place recognition in indoors for a humanoid robot. It presented some applications that consider indoor scenarios, such as the system developed in [1] this article that uses effective and redesigned convolutional neural networks to recognize objects and scenes in indoor environments. With the goal of helping blind and visually impaired people explore and navigate indoor environments, the system produces state-of-the-art results on indoor datasets. As shown in Fig. 1, information from completely connected parallel networks is added to the model to help it pay more attention to the information contained in interior picture images. The features extracted by the convolutional neural networks are input to a fully connected layer for each class. Then, the features processed by these networks are input to a fully connected layer to obtain indoor environment classification results. After training the model with the training set, the validation set is used to verify the accuracy of the model. 2.2

Local Feature Extraction

The SIFT feature detector is a traditional method for extracting distinctive features from an image that do not vary with scale and rotation changes. After identifying the location and scale of interest points by applying Gaussian difference filters, highly stable interest points are selected as key points. To refine locations with subpixel precision and assign orientation to represent rotation-invariant key points, the dominant orientation is assigned to each key point based on the

282

L. Hernando R´ıos Gonz´ alez et al.

Fig. 1. The Model Architecture CNN

location of the local image gradient. Finally, each key point is described using local gradients to capture significant levels of local shape distortion and illumination change as a 128-dimensional vector representing gradient orientations in the local neighborhood. It has been shown that the SIFT descriptor is resistant to illumination changes, noise, and minor distortions. BoVW The bag of visual words (BoVW) approach is a common technique for image classification and retrieval. It represents images as “bags” of “visual words” that can be compared quantitatively. Local functions such as SIFT are first extracted from the training images and clustered to form a vocabulary of visual words, where each cluster center is a visual word. For a new image, each local feature is assigned to its nearest visual word to obtain a histogram of visual words for the image. Then, images are compared through visual word histograms using a metric such as chi-square distance. Local descriptors, such as SIFT, are extracted using unsupervised clustering techniques, such as k-Means, where each visual word is represented by the centroid of each group. The number of times each visual word appears in a given image is used to represent the image, generating a BoVW histogram. A Support Vector Machine (SVM) is trained for each object category based on the visual word histogram using the histograms corresponding to the training pictures with the objects in that category as positive examples and the histograms of the other categories as negative examples. The visual word histogram of a picture containing an unidentified item to categorize is obtained and sent to the available SVMs. The image is then categorized into the SVM object category with the greatest score [16]. VLAD (Vector Locally Aggregated Descriptor) The goal of VLAD (Vector of Locally Aggregated Descriptors) is to compact the information extracted by interest points from K-Means into a vector that represents the analyzed image. Unlike the Fisher vector-based descriptor, VLAD does not use Gaussian mixture models to cluster interest points and build a dictionary of words, instead it uses K-Means clustering and the centroid of each of its clusters or groupings as a parameter for assignment and histogram construction. In the case of VLAD, the variation measure will be the distance between point and centroid maintaining the sign of change, in order to know the direction with respect to the centroid. By summing for each group of points the calculated distances, a

Evaluation of Feature Descriptors for Scene Classification

283

one-dimensional vector is obtained for each grouping. The concatenation of all the resulting distance vectors generates the VLAD local representation vector. The Vector of Locally Aggregated Descriptors (VLAD) is an image representation that adds local feature vectors with their assigned cluster centers. Given the local feature vectors pi and cluster centers ck , the VLAD representation V is a K × D matrix where K is the number of clusters and D is the dimension of the feature vector. Each element V (j, k) is calculated as the sum of the residuals of the features assigned to group j: V (j, k) =

N 

(pi (j) − ck (j))

(2)

i=1

After calculating the VLAD matrix V, it is converted into a vector and the L2 norm is used to obtain the final VLAD image representation. VLAD aims to capture the descriptive information that is lost when features are simply quantized into visual words by aggregating residual distances to group centers [2]. Fisher vector The Fisher vector (FV) is a feature encoding technique that represents local features by their deviation from a Gaussian mixture model (GMM). A GMM is fitted to the local features of the training images. For a new image, each local feature is assigned to the gradient of the log likelihood with respect to the means and covariances of each component, which captures how the feature deviates from the expected distribution. The sum of these gradients over all features gives the Fisher vector, which is then normalized. It has been shown that FV achieves state-of-the-art results for image classification and retrieval, but is more complex to compute than simpler encoding approaches [13].

3

Experimental Evaluation and Results

Learning how humans see and understand their environment and mimicking this ability in mobile robots by using cutting-edge artificial intelligence techniques to scene recognition has been a goal of cutting-edge research.Mobile robots using cameras and computer vision software can use artificial intelligence to recognize and categorize objects in their working environment.The tests are designed to mimic various complex situations that mobile robots encounter on a daily basis, such as the interaction between the robot and the environment it perceives while moving, specific scenes or scenes with objects from various perspectives, and changes in the lighting conditions of various scenes. The image data set was subjected to numerous meticulous manipulations that allow evaluation of the performance of descriptors BoVW, Fisher, VLAD, GIST, and CNN in challenging recognition tasks under conditions that simulate the complex changes and variations that mobile robots in dynamic environments must inevitably face. Changes in perspective, the removal of set pieces, and the use of a few props that belong to the scenes are among the manipulations that have been made, both in training images and test images.

284

3.1

L. Hernando R´ıos Gonz´ alez et al.

Database

The data used in the proposed structure were meticulously collected to test the performance of visual perception techniques under conditions similar to those experienced by mobile autonomous robots in the actual world. First, a basis of original images was built by taking a large number of photographs within the buildings, including internal spaces like salas and passageways. All scenes from the underlying material depict complex environments where objects are subject to significant transformations, such as adjustments to lighting, perspective, and size. Manual data collection was done in stages to ensure the thoroughness and accuracy of the annotations. First, controlled lighting conditions were used to take the photographs. In order to simulate the practical challenges that mobile robots with optical sensors constantly face, the collection of data was then subjected to additional manipulations including zooming, perspective changes, and lighting variations. The proprietary image base contains a total of 500 photographs, categorized into Room One, Room Two, Room Three, Corridor 0ne, and Corridor Two scenes as shown in the Fig. 2, 250 images were taken of long, narrow corridors subject to high traffic, with multiple lighting including fluorescent and LED lights, as well as natural light. The images capture the corridors in different conditions (occupied and empty) and from multiple perspectives (front, side and oblique). 375 scenes of complex rooms, such as family dining rooms, living rooms and television rooms, were photographed, illuminated with natural and artificial light. These scenes recreate domestic environments with a wide variety of objects and occupancy levels. In addition, close-ups were taken of the most relevant objects and areas of interaction (sofas, tables, coffee tables, etc.) to provide richer annotations. 3.2

Experimental

Next, a series of tests is performed with the own database to analyze the performance of the BoVW, VLAD, FISHER, GIST descriptors and CNN. When PCA and RELIEF are applied. To perform the test, a database consisting of 5 classes of indoor environment images was used;ROOM ONE, ROOM TWO, ROOM THREE, CORRIDOR ONE and CORRIDOR TWO. The following is a brief description of each of the cases. Case 1. Comparison of results of all descriptors using training images—cropped or incomplete scenes. Each image of each class of the training base is applied the IMPLIT function to divide it into 4 parts and obtain cropped images. Test images—complete scenes. Complete images of the OWN BASE. Case 2. Comparison of results of all descriptors using training images—cropped or incomplete scenes. Each image of each class of the training base is applied the IMPLIT function to divide it into 4 parts and obtain cropped images. Test

Evaluation of Feature Descriptors for Scene Classification

285

images—complete scenes. Complete images of the OWN BASE modified by changes in rotation and perspective. Case 3. Comparison of results of all descriptors using training images—cropped or incomplete scenes. Each image of each class of the training base is applied the IMPLIT function to divide it into 4 parts and obtain cropped images. Test images—complete scenes. Cropped images according to the position and specific dimensions in the cropping rectangle, which seek to highlight objects with high semantic information. Case 4. Comparison of results of all descriptors using training images—complete scenes. Test images—individual objects or parts or regions of the scenes. The main feature of this database is that the training images contain all the objects of the scene and the test images contain specific objects, seen from different points of view or isolated. Table 1. Summary of the performance of the descriptors applied to the proprietary image database Cases/Descriptors BoVW

VLAD

FISHER GIST CNN

Cases 1

0.68

0.72

0.66

0.52

0.81

Cases 2

0.84

0.79

0.79

0.61

0.97

Cases 3

0.69

0.74

0.52

0.56

0.82

Cases 4

0.77

0.71

0.74

0.63

0.90

Average

0.77

0.74

0.67

0.60

0.88

The following Table 1 shows the performance of the different descriptors for the 4 test cases performed. From the different cases it is shown that some descriptors present better results than others in carrying out the tests.For cases 2 and 4, the BoVW descriptor proved to be the most efficient. For case 1, the VLAD descriptor presents better performance. The descriptor that presented the lowest performance for all cases was the GIST descriptor. The descriptor that presented the best performance for all cases was the BoVW descriptor.One of the main characteristics of the own image base was the similarity of Scenario 4 CORRIDOR ON and Scenario 5 CORRIDOR TWO classes. In the different tests performed, it was shown that the classifiers confused many of the images of these classes. Table 2. Scene Classification Performance on MIT Indoor 67 Cases/Descriptors BoVW Cases 1

VLAD

FISHER GIST

0.6318 [23] 0.6612 [7] –

CNN

0.261 [15] 0.7638

286

L. Hernando R´ıos Gonz´ alez et al.

When the visual recognition method is validated on a state-of-the-art data set like MIT Indoor 67, not only are the inherent difficulties of the task seen, but also the areas of opportunity for the data set itself to advance the state of the technique.Validating in her reveals the limitations of a chosen methodology as well as the data base’s ability to accurately portray the complexity of visual recognition in real-world circumstances.as seen in the Table 2. 3.3

Discussion

Although the descriptors allow for highlighting some aspects of the picture, this is insufficient to provide a good classification prior to recognition. When the characteristics of the images’ bases change due to factors like occlusion, illumination, rotation, and perspective of the picture, all the descriptors under analysis exhibit deficiencies in classification.These changes bring about variability in the visual representations, which the classifiers find difficult to manage due to the vast amounts of information that the descriptors manage. In general, this situation is resolved by optimizing the data, removing arbitrary values through analysis of key components, and providing meticulous data organization to make separation easier.One of the solutions adopted to increase classification performance is the implementation of global descriptor techniques such as BoVW, VLAD, FISHER, but these techniques, although they improve data characteristics such as dimensionality, since they allow generating dictionaries or visual bags of words, also present a low performance in classification since they do not allow the generality or universality of the groupings. At present, convolutional neural networks—CNN are the most used technique for scene recognition and their potential lies in the chain of neural blocks that allow generalization of recognition patterns from beginning to end. Given that local descriptors have their greatest strength in describing low-level feature descriptions, and global descriptors in describing representations, hybrid models present better results in recognition tasks. A test to perform is to take the best performing descriptor from each case and create a hybrid system with a CNN, to obtain a system that captures both local feature information and global feature information in the recognition task.

4

Conclusion

The performance of the local and global image descriptors SIFT, BoVW, VLAD, FISHER, GIST, and CNN was examined in this article in order to recognize complex interior environment scenes that attempt to mimic the many locations through which a mobile robot can go. The results of the tests conducted on the four study cases demonstrated the effectiveness of the artisanal descriptors when techniques are used to enhance the data input into the classifier.Although it is known that these methods are reliable and provide good precision for image recognition, their performance significantly decreased when scenes were captured from slightly different angles. Even with minor rotations or perspective and scale

Evaluation of Feature Descriptors for Scene Classification

287

shifts in relation to the points of view that the clasificators were formed from. According to these experiments, La CNN was the method that produced the best results, but unlike the other methods, it has a high computational cost.Of the handcrafted methods analyzed, the BoVW method is the most resistant to changes in viewpoint and background noise, making it the most appropriate for recognizing complex scenes. Table 2 of the article shows the performance of the implemented descriptors using a state of the art image base as MIT indoor 67. From this table we conclude that the handcrafted descriptors and the implemented CNN present acceptable performance with state of the art image bases .In future work, our goal is to implement a local and global feature representation based on CNN networks, using the test base of the different case studies.The final scope of application is that a robot may be able to recognize the place where it is located and provide a first collection of training images of the different scenarios where it interacts, e.g. different indoor and outdoor environments. Acknowledgements. This research has been supported by the project “COordinated intelligent Services for Adaptive Smart areaS (COSASS), Reference: PID2021123673OB-C33, financed by MCIN/AEI/10.13039/501100011033/FEDER, UE.

References 1. Afif, M., Ayachi, R., Said, Y., Atri, M.: Deep learning based application for indoor scene recognition. Neural Process. Lett. 51, 2827–2837 (2020) 2. Arandjelovic, R., Gronat, P., Torii, A., Pajdla, T., Sivic, J.: Netvlad: Cnn architecture for weakly supervised place recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 5297–5307 (2016) 3. Bay, H., Tuytelaars, T., Van Gool, L.: Surf: Speeded up robust features. Lect. Notes Comput. Sci. 3951, 404–417 (2006) 4. Bosch, A., Zisserman, A., Munoz, X.: Image classification using random forests and ferns. In: 2007 IEEE 11th International Conference on Computer Vision, pp. 1–8. IEEE (2007) 5. Csurka, G., Dance, C., Fan, L., Willamowski, J., Bray, C.: Visual categorization with bags of keypoints. In: Workshop on Statistical Learning in Computer Vision, ECCV, vol. 1, pp. 1–2. Prague (2004) 6. Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), vol. 1, pp. 886–893. IEEE (2005) 7. Gong, Y., Wang, L., Guo, R., Lazebnik, S.: Multi-scale orderless pooling of deep convolutional activation features. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014. LNCS, vol. 8695, pp. 392–407. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-10584-0 26 8. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016) 9. J´egou, H., Perronnin, F., Douze, M., S´ anchez, J., P´erez, P., Schmid, C.: Aggregating local image descriptors into compact codes. IEEE Trans. Pattern Anal. Mach. Intell. 34(9), 1704–1716 (2011)

288

L. Hernando R´ıos Gonz´ alez et al.

10. Jurie, F., Triggs, B.: Creating efficient codebooks for visual recognition. In: Tenth IEEE International Conference on Computer Vision (ICCV’05), vol. 1, pp. 604– 610. IEEE (2005) 11. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521(7553), 436–444 (2015) 12. Lowe, D.G.: Distinctive image features from scale-invariant keypoints. Int. J. Comput. Vis. 60, 91–110 (2004) 13. Murino, V., Puppo, E.: Image Analysis and Processing-ICIAP 2015: 18th International Conference, Genoa, Italy, Proceedings, Part I, vol. 9279. Springer (2015) 14. Oliva, A., Torralba, A.: Modeling the shape of the scene: A holistic representation of the spatial envelope. Int. J. Comput. Vis. 42, 145–175 (2001) 15. Quattoni, A., Torralba, A.: Recognizing indoor scenes. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp. 413–420. IEEE (2009) 16. Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al.: Imagenet large scale visual recognition challenge. Int. J. Comput. Vis. 115, 211–252 (2015) 17. S´ anchez, J., Perronnin, F., Mensink, T., Verbeek, J.: Image classification with the fisher vector: Theory and practice. Int. J. Comput. Vis. 105, 222–245 (2013) 18. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., Rabinovich, A.: Going deeper with convolutions. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1–9 (2015) 19. van Gemert, J.C., Geusebroek, J.-M., Veenman, C.J., Smeulders, A.W.M.: Kernel codebooks for scene categorization. In: Forsyth, D., Torr, P., Zisserman, A. (eds.) ECCV 2008. LNCS, vol. 5304, pp. 696–709. Springer, Heidelberg (2008). https:// doi.org/10.1007/978-3-540-88690-7 52 20. Vogel, J., Schiele, B.: A semantic typicality measure for natural scene categorization. In: Rasmussen, C.E., B¨ ulthoff, H.H., Sch¨ olkopf, B., Giese, M.A. (eds.) DAGM 2004. LNCS, vol. 3175, pp. 195–203. Springer, Heidelberg (2004). https://doi.org/ 10.1007/978-3-540-28649-3 24 21. Wei, X., Phung, S.L., Bouzerdoum, A.: Visual descriptors for scene categorization: experimental evaluation. Artif. Intell. Rev. 45, 333–368 (2016) 22. Wozniak, P., Afrisal, H., Esparza, R.G., Kwolek, B.: Scene recognition for indoor localization of mobile robots using deep CNN. In: Chmielewski, L.J., Kozera, R., Orlowski, A., Wojciechowski, K., Bruckstein, A.M., Petkov, N. (eds.) ICCVG 2018. LNCS, vol. 11114, pp. 137–147. Springer, Cham (2018). https://doi.org/10.1007/ 978-3-030-00692-1 13 23. Wu, R., Wang, B., Wang, W., Yu, Y.: Harvesting discriminative meta objects with deep cnn features for scene classification. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 1287–1295 (2015) 24. Xie, L., Lee, F., Liu, L., Kotani, K., Chen, Q.: Scene recognition: A comprehensive survey. Pattern Recognit. 102, 107205 (2020) 25. Zeiler, M.D., Fergus, R.: Visualizing and understanding convolutional networks. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014. LNCS, vol. 8689, pp. 818–833. Springer, Cham (2014). https://doi.org/10.1007/978-3-31910590-1 53 26. Zhou, X., Zhuang, X., Tang, H., Hasegawa-Johnson, M., Huang, T.S.: Novel gaussianized vector representation for improved natural scene categorization. Pattern Recognit. Lett. 31(8), 702–708 (2010)

Influence of Segmentation Schemes on the Interpretability of Functional Connectivity in Mild Cognitive Impairment Isabel Echeverri-Ocampo1 , Karen Ardila1 , José Molina-Mateo2 , Jorge Iván Padilla-Buriticá3 , Belarmino Segura-Giraldo4 , Hector Carceller5 , Ernesto A. Barceló-Martinez6 , and Maria de la Iglesia-Vaya5(B) 1 Departamento de Electrónica y Automatización-Universidad Autónoma de Manizales, Caldas,

Colombia {Colombiaisabelcecheverrio,karen.ardilal}@autonoma.edu.co 2 Center for Biomaterials and Tissue Engineering, Universitat Politècnica de València, Valencia, Spain [email protected] 3 AMYSOD LabParque i, CM&P Research Group, Instituto Tecnológico Metropolitano ITM, CL 73 No. 76 A 354, Medellín, Colombia 4 Plasma Physics Laboratory, Universidad Nacional de Colombia Sede Manizales, Campus la Nubia, Manizales, Colombia [email protected] 5 Biomedical Imaging Unit FISABIO-CIPF, Fundación Para el Fomento de la Inves-tigación Sanitaria y Biomédica de la Comunidad Valenciana, 46012 Valencia, Spain [email protected], [email protected] 6 Neurologist, Instituto Colombiano de Neuropedagogía, Universidad de la Costa, Facultad de Psicología, Barranquilla, Colombia [email protected]

Abstract. The segmentation of electroencephalography (EEG) signals is becoming increasingly important for the monitoring of neurological behaviors in clinical practice. This paper presents the influence of wild binary segmentation -change point detection technique on the interpretability of functional connectivity. The EEG test was recorded during visual odd-ball tasks to 30 subjects (19 females; mean age of 70.63 years old, age range 61–79 years old, 11 males; mean age of 70.36 years old, and age range 63–81 years old) 14 with mild cognitive impairment and 16 control group subjects. The method used to find the window segments is taken from R CRAN packages; Wild Binary Segmentation. Results show that the Wild Binary Segmentation (WBS) method had an adequate performance in signal segmentation at beta and alpha band, in addition to ob-serving agreement between the detected windows and the status quo of the beta frequency band in the brain. Keywords: EEG · Change point detection · Mild cognitive impairment

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 289–297, 2023. https://doi.org/10.1007/978-3-031-36957-5_25

290

I. Echeverri-Ocampo et al.

1 Introduction Mild cognitive Impairment (MCI) is a clinical condition characterized by cognitive decline that exceeds normal age-related changes but does not meet the criteria for dementia. It is a heterogeneous condition with various underlying pathologies, including Alzheimer’s disease (AD), vascular dementia, and lewy body dementia [1]. Identify biomarkers that can accurately diagnose MCI and differentiate it from normal aging and dementia is crucial for early intervention and management [2]. Functional connectivity (FC) analysis is a popular tool for researching brain function in MCI. FC measures the statistical dependence between brain regions, providing insights into the functional organization of the brain. However, the interpretation of FC can be confounded by many factors, including the choice of segmentation method [3]. In this study, we investigate the influence of segmentation on the interpretability of FC in MCI using Wild Binary Segmentation (WBS) method [4]. WBS is a datadriven segmentation method that identifies abrupt changes in the signal, providing a more detailed segmentation of the brain that conventional method [5]. Our results demonstrate that WBS segmentation significantly improves the interpretability of FC in MCI. We show that WBS-based segmentation reveals distinct FC patterns that are not observed with conventional segmentation methods, suggesting that WBS may be a valuable tool for identifying biomarkers of MCI.

2 Methods This study focused on applying a method of change point detection in EEG time series. Before reaching the method, signal processing was performed to eliminate the noise. The CPD was carried out in alpha and beta band of the time series. 2.1 EEG Data Preprocessing The dataset includes 30 adults over the age of 60–14 patients diagnosed with MCI and 16 control healthy patients (age range of 61 to 81 years old, with an average age of 70.53 years). This dataset used a visual three-stimulus oddball paradigm, see Fig. 1. As a first preprocessing stage, we proceed to the filtering using digital band-pass filters, specifically FIR (finite impulse response) filters applying the FIRWIN form, which consists of utilizing the window method. These filters use a finite number of numerator coefficients, can be adjusted to zero phase without extra computations, are easy to manage, have a well-defined passband, and can be converted to minimum-phase. As a result, FIR filters are recommended for the majority of electrophysiological data processing applications [6]. Therefore, in order to split the signal into each brain rhythm the following frequencies are considered: Theta 4–8 Hz, Alpha 8–12 Hz, Beta 12–35 Hz and Gamma 35–40 Hz [7–9]. Nevertheless, from the observation and analysis of the spectrogram of the raw signal, it is considered pertinent to maintain only the alpha and beta bands. Additionally, we used ICA (Independent Components Analysis) which consists in a signal processing method to separate independent sources linearly mixed in several

Influence of Segmentation Schemes on the Interpretability of Functional Connectivity

291

Fig. 1. Schematic of the visual oddball paradigm

sensors, thanks to the statistical independence and the projection of these, to provide a filtering of the artifacts and ocular noise (eye blink) from the signal [10]. The method employed to apply ICA is infomax, due to its performance in terms of mutual information reduction as well as the remaining pairwise mutual information against the per-cent of near-dipolar components, having more independence and consequently more biological components. In addition, it tends to return more dipolar components, being more efficient [11]. 2.2 EEG Change Point Detection Analysis In this paper, Principal Components Analysis (PCA) is used in order to reduce the dimension of the EEG signals. In this context, PCA can search the maximum variation from the input higher dimensions. Consequently, by reducing the channels of interest. The EEG signal is transformed into a specific signal using PCA. According to our work, the frontal group is made up of the following channels: F7, F3, FZ, F4, F8. We implemented PCA on these 5 channels and took the first PCA. We divide the electrodes into zones frontal, temporal, parietal and central to implement PCA to decrease the computational demand of our change point analysis. In order to overcome the nonstationarity present in brain activity, change point detection schemes must be considered for the analysis of the EEG signal; we perform Wild Binary Segmentation (WBS) to mitigate the non-stationarity of the signal. This method uses the idea of computing Cumulative Sum (CUSUM) from randomly generated ranges considering the largest CUSUM to the first change point position [12]. 2.3 Functional Connectivity Functional connectivity estimates are highly dependent on brain activities over time; we used EEG as a technique for capturing the dynamic and quick neural response, due to its high temporal resolution [13].

292

I. Echeverri-Ocampo et al.

WPLI is a measure of phase synchronization to functional connectivity and was created to obtaining reliable phase synchronization estimates that are invariant in the presence of common sources such as volume conduction [14, 15], and is calculated as follows: WPLI =

|E[Im(Sxy)]| E[|Im(Sxy)|]

(1)

where E[] denotes the hope operator applied on the epochs, Im() represents the imaginary part and Sxy is the cross spectrum between x and y.

3 Results and Discussions The average locations of the first ten change-points of the PCA signals were 1000 ms and 2000 ms with the three change points methods. The length of the segments was verified by the augmented Dickey Fuller test [16]. The examined signals appeared nonstationary. The test Kolmogorov–Smirnov (KS) applied to the methods windows in alpha band, because of p-values of MCI subjects using KS in alpha band with WBS is 0.02004 and beta band is 0.0024. The window distributions of WBS method confirmed that EEG is a non-normal distribution. Figure 2 shows the distribution of the segments.

Fig. 2. Estimation of connectivity through change point methods for healthy patients and MCI. a 1st row corresponds to the EEG signal with PCA method. b 2nd row shows the change point detection in alpha and beta band. c 3rd row is the adjacency matrix. d 4th row thresholding percentage.

Influence of Segmentation Schemes on the Interpretability of Functional Connectivity

293

3.1 EEG Segmentation Method Signal Analysis The results according to the alpha and beta band are adjusted with the fact that alpha band power is negatively correlated with brain activity. On the other hand, in beta band activity we observe the opposite effect. Since the beta, band activity is in accordance with the “status quo maintenance” hypothesis, which states that the power of this band is lower when we expect cognitive process to occur that change with status quo [17]. 3.1.1 Alpha Band Functional Connectivity Analysis The alpha band is characterized by being a fundamental for controlling cognitive processes, such as attention, perception, functional inhibition and working memory [18]. According to these, we found lower activity in people with MCI in tempo-parietal areas [19], compared to healthy ones; this finding is more appreciable in the target stimulus when the brain is processing the information of oddball protocol. However, in the other areas we did not find any differences between the two groups during the visual oddball protocol. The alpha-band behavior at the beginning of a stimulus-locked EEG epoch and gradually declines as the change-point moves to the right [20]. Table 1. Table of Critical Value in alpha band for WBS method for CPD in EEG in MCI subjects Change point

Time (ms)|

p-Values

Effect size

CP1

200–400

0.001

0.92

CP2

450–600

0.012

0.71

CP3

800–1000

0.008

0.81

CP4

1200–1400

0.021

0.62

CP5

1600–1800

0.003

0.87

CP6

2000–2200

0.014

0.74

CP7

2400–2600

0.007

0.83

CP8

3200–3400

0.016

0.71

CP9

3600–3800

0.005

0.89

In Table 1 showed that there were two significant change point detection in the EEG time series, indicating shifts in brain activity during the target task. The first change point occurred within 500 ms of the task, suggesting an early adjustment of the cognitive process. The second change point occurred between 900 and 1000 ms, indicating a sustained cognitive effort to maintain the visual information in working memory. The significant change point in alpha activity indicates there are differences at time in intervals between MCI subjects. The effect size in the table ranges from 0.53 to 0.92, indicating that the observed differences in alpha activity is moderate to large. The p-values from 0.001 to 0.032, indicating the observed differences are statistically significant.

294

I. Echeverri-Ocampo et al.

3.1.2 Beta Band Functional Connectivity Analysis Oscillatory development interrelated to the EEG frequencies bands play a relevant role in functional connectivity during cognitive progress. The present study analyzes agerelated functional connectivity during visual tasks. The results indicate that the aging differences appear during visual stimuli in the beta band [21]. Activity we observe the opposite effect. Since the beta band activity is in accordance with the “status quo maintenance” hypothesis, which states that the power of this band is lower when we expect cognitive process to occur that change with status quo [17]. Table 2. Table of Critical Value in beta band for WBS method for CPD in EEG in MCI subjects Change point

Time (ms)

p-Values

Effect size

CP1

500

0.023

– 0.58

CP2

1200

0.005

– 0.72

CP3

2000

0.018

– 0.62

The status quo hypothesis was that the power of beta band would be lower when cognitive process was expected, so a decrease in beta power would indicate the presence of a change point. Table 2 shows three significant change points identified at 500 ms, 1200ms, and 2000 ms. The p-values for each one indicates the probability of observing a change as extreme as the one detected under the null hypothesis of no change is low, suggesting that the change points are statistically significant. 3.2 Functional Connectivity The analysis revealed that individuals with mild cognitive impairment exhibited reduced functional connectivity in the alpha and beta bands, particularly in the frontal and parietal regions of the brain. The change point detection analysis showed that there was a significant change in functional connectivity patterns in the alpha band in individuals with mild cognitive impairment compared to those with normal cognition [23]. Similarly, in the beta band, the change point detection analysis revealed a significant reduction in functional connectivity in individuals with mild cognitive impairment compared to those with normal cognition. This reduction was observed in the frontal and temporal regions of the brain, see Fig. 3, indicating a disrupted functional connectivity in these regions in individuals with mild cognitive impairment [24]. Overall, these results suggest that functional connectivity in the alpha and beta bands of the electroencephalogram can be used as a biomarker for mild cognitive impairment, as there is a significant reduction in functional connectivity patterns in individuals with mild cognitive impairment compared to those with normal cognition. These findings have important implications for the early detection and treatment of mild cognitive impairment and for the development of more effective therapeutic strategies for this condition.

Influence of Segmentation Schemes on the Interpretability of Functional Connectivity

295

Fig. 3. Brain maps of phase synchronization index. The lines connecting electrode sites indicate significantly higher.

4 Conclusions This paper has presented the influence of one change-point detection methods in EEG recording of middle impairment cognitive signals. This method uses change point detection strategy evaluating the temporal of alpha and beta band. Because there is a meager amount of difference in the change point analysis, we research the dependence of the task-induced segments, relying on the p-value that is implemented as a statistical analysis. WBS showed an effective performance on both bands. In addition, the analysis indicates that there are differences between non-target and target stimuli during the attention task. The alpha band activity is stronger at the start of the stimulus-locked EEG epoch and gradually diminishes as the change-point advances to the right. This is consistent with the “inhibition timing” hypothesis, according to which alpha band power is negatively correlated with brain activity; it is higher during the pre-stimulus interval but decreases as the frontal cortex becomes involved in task-related decision-making regarding the response to the target trials. Overall, these results suggest that change point detection in beta band of EEG time series can be useful tool for investigating changes in cognitive process and may help to shed light on the underlying neural mechanisms involved in cognitive function. In future work, incorporating features unique to this population; we aim to improve the senility and specificity of change point detection methods in identifying early signs

296

I. Echeverri-Ocampo et al.

of cognitive decline. While there are, still challenges to be addressed, such us the need for larger and more diverse dataset to validate the algorithm, we believe that this future work has the potential to contribute to the field of cognitive neuroscience and improve clinical practice in the early detection and management of mild cognitive impairment.

References 1. Miller, K.L., Pedelty, L., Testai, F.D.: The relationship between heart disease and cognitive impairment. Handb. Clin. Neurol. 177, 377–391 (2021) 2. Jankovic, J., Mazziotta, J., Pomeroy, S.: Bradley and Daroff’s Neurology in Clinical Practice, 2-Volume Set, 8th ed., vol. 2. Elsevier (2021) 3. Allen, E.A., Damaraju, E., Eichele, T., Wu, L., Calhoun, V.D.: EEG signatures of dynamic functional network connectivity states. Brain Topogr. 31(1), 101 (2018) 4. P. Fryzlewicz, “Wild binary segmentation for multiple change-point detection,” vol. 42, no. 6, pp. 2243–2281 (2014). https://doi.org/10.1214/14-AOS1245 5. Gaur, P., Gupta, H., Chowdhury, A., McCreadie, K., Pachori, R. B., Wang, H.: A sliding window common spatial pattern for enhancing motor imagery classification in EEG-BCI. IEEE Trans. Instrum. Meas. 70 (2021) 6. Widmann, A., Schröger, E., Maess, B.: Digital filter design for electrophysiological data – a practical approach. J. Neurosci. Methods 250, 34–46 (2015) 7. Abhang, P.A., Gawali, B.W., Mehrotra, S.C.: Introduction to EEG- and Speech-Based Emotion Recognition. Elsevier Inc. (2016) 8. Kropotov, J. D.: Functional neuromarkers for psychiatry: applications for diagnosis and treatment. Elsevier Inc. (2016) 9. Binder, M.D., Hirokawa, N., Windhorst, U. (eds.): Encyclopedia of Neuroscience. Springer, Heidelberg (2009). https://doi.org/10.1007/978-3-540-29678-2 10. Delorme, A., Sejnowski, T., Makeig, S.: Enhanced detection of artifacts in EEG data using higher-order statistics and independent component analysis. Neuroimage 34(4), 1443–1449 (2007) 11. Delorme, A., Palmer, J., Onton, J., Oostenveld, R., Makeig, S.: Independent EEG sources are dipolar. PLoS ONE 7(2), e30135 (2012) 12. Caplette, L., Ince, R.A.A., Jerbi, K., Gosselin, F.: Disentangling presentation and processing times in the brain. Neuroimage 218, 116994 (2020) 13. Ismail, L.E., Karwowski, W.: A graph theory-based modeling of functional brain connectivity based on EEG: A systematic review in the context of neuroergonomics. IEEE Access 8, 155103–155135 (2020) 14. Imperatori, L.S. et al.: EEG functional connectivity metrics wPLI and wSMI account for distinct types of brain functional interactions. Sci. Reports 2019 91 9(1), 1–15 (2019) 15. Vinck, M., Oostenveld, R., Van Wingerden, M., Battaglia, F., Pennartz, C.M.A.: An improved index of phase-synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size bias. Neuroimage 55(4), 1548–1565 (2011) 16. Dickey, D.A., Fuller, W.A.: Distribution of the estimators for autoregressive time series with a unit root. J. Am. Stat. Assoc. 74(366), 427 (1979) 17. Barone, J., Rossiter, H.E.: Understanding the role of sensorimotor beta oscillations. Front. Syst. Neurosci. 15, 51 (2021) ´ ci´c-Blake, B.: Alpha power and func18. Lejko, N., Larabi, D.I., Herrmann, C.S., Aleman, A., Curˇ tional connectivity in cognitive decline: a systematic review and meta-analysis. J. Alzheimers. Dis. 78(3), 1047–1088 (2020)

Influence of Segmentation Schemes on the Interpretability of Functional Connectivity

297

19. Clayton, M.S., Yeung, N., Cohen Kadosh, R.: The many characters of visual alpha oscillations. Eur. J. Neurosci. 48(7), 2498–2508 (2018) 20. Klimesch, W.: Alpha-band oscillations, attention, and controlled access to stored information. Trends Cogn. Sci. 16(12), 606–617 (2012) 21. Fodor, Z., Horváth, A., Hidasi, Z., Gouw, A.A., Stam, C.J., Csukly, G.: EEG alpha and beta band functional connectivity and network structure mark hub overload in mild cognitive impairment during memory maintenance. Front. Aging Neurosci. 13, 668 (2021) 22. Wang, L. et al. (2017) Beta-band functional connectivity influences audiovisual integration in older age: An EEG study. Front. Aging Neurosci. 9(AUG), 239 (2017) 23. Hird, M.A., Churchill, N.W., Fischer, C.E., Naglie, G., Graham, S.J., Schweizer, T.A. (2018) Altered functional brain connectivity in mild cognitive impairment during a cognitively complex car following task. Geriatr. (Basel, Switzerland) 3(2) (2018) 24. Gurja, J.P., Muthukrishnan, S.P., Tripathi, M., Sharma, R.: Reduced resting-state cortical alpha connectivity reflects distinct functional brain dysconnectivity in Alzheimer’s disease and mild cognitive impairment. Brain Connect. 12(2), 134–145 (2021)

The Content Based Misinformation Detection for Gujarati Language Uttam Chauhan1 , Vinay Sheth1 , Vishvesh Trivedi1 , Chintan Bhatt2(B) , and Juan Manuel Corchado3,4,5 1

2

Vishwakarma Government Engineering College, Ahmedabad, India ug [email protected] Department of Computer Science and Engineering, School of Technology, Pandit Deendayal Energy University, Gandhinagar, Gujarat, India [email protected] 3 BISITE Research Group, University of Salamanca, 37007 Salamanca, Spain [email protected] 4 Air Institute, IoT Digital Innovation Hub, 37188 Salamanca, Spain 5 Faculty of Engineering, Department of Electronics, Information and Communication, Osaka Institute of Technology, Osaka 535-8585, Japan

Abstract. Misinformation detection (MID) has recently gained attention as a research question. We discover that many efforts have been made in response to the innovative research issues and MID research approaches as an empowering and developing rapidly research field. Misinformation typically spreads faster, profound, and wider in social networks. It is crucial to identify disinformation on social media since people are less able to distinguish between true information and false information due to an abundance of information and limited attention. Exploration of misinformation detection received little attention from the computational NLP research community. In this paper, we propose a novel architecture for detecting misinformation in Gujarati text. In the architecture, the significance of domain experts, crowd intelligence, and factchecking website interaction has been depicted. Additionally, we define certain special Gujarati language traits for the early detection of counterfeit news. We have built a classifier incorporating domain experts to identify fake news and present outcomes for the automatic fake news detection and its class. We showed the findings of the experiments performed on the news article corpus that we built by fetching the articles from newspaper websites and applications. We found that stochastic Gradient Descent and NuSVC achieved the accuracy 88% using TFIDF and CV respectively, which is the highest among all models we experimented. Keywords: Misinformation Detection · Fake News Detection · Crowd Intelligence · Social Network Data Analysis · Text Classification

c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 298–309, 2023. https://doi.org/10.1007/978-3-031-36957-5_26

The Content Based Misinformation Detection for Gujarati Language

1

299

Introduction

Social media’s transparency and promptness has allowed misinformation, such as rumors, spam, and fake news, to spread rapidly. Users of Twitter, Instagram, Snapchat, and WhatsApp can interact with online content, strangers, and their peers. Social media has developed and gained in importance among people, especially young people. Youth are currently more proactively engaging in social initiatives, protests, and news dissemination in general thanks to social media. A study conducted by CNN research, 62% of Americans get their information from social media, and 74% get their news via email or social media updates. The use of this disinformation is greatly used to manipulate the results of various elections in the world. Information that is erroneous or misleading is misinformation, by definition. False rumours, insults, and practical jokes are instances of misinformation, but malevolent content like hoaxes, spear-phishing, and computational propaganda are examples of more deliberate disinformation. In addition to leading people to adopt false beliefs and influencing how they perceive to the truth, the ubiquitous misinformation also undermines the trustworthiness of the entire information ecosystem. What’s worse is that with the quick proliferation of social networking platforms, the foreseeable future of online deceit, such as rumours, fake news, and so on, will go far beyond text to incredibly potent and deceptive information resources, such as photographs, videos, and audios on a large scale [5]. Facebook and other online social media sites make it easier to maintain and develop relationships with others. It helps users identify themselves by enabling them to create profiles and share information with other members through text, pictures, and photos. Facebook, Twitter, Instagram, WhatsApp, LinkedIn, WeChat, Snapchat, and Foursquare are just some of the most widely used social media platforms. The frequency of sharing content on online social networks has increased along with the popularity of these platforms. The shift in behaviour for such consumption can be attributed to a variety of factors. Especially in comparison to newspapers or other traditional news organisations, social media sites demand less time and money to deliver the same content. Sharing material in the form of blogs, posts, or videos with friends or other users is also made simpler. Because of this, authors and publishers can more easily post their contents as articles in group settings. Since 2017, there has been an upsurge in social media usage of approximately 13% globally. Through social networking platforms, news information has been shared and created in the form of posts, blogs, articles, photographs, videos, etc. The emergence of social media opened up a platform for the dissemination of false information online [2]. Figure 1 shows how misinformation is havoc creating in social media. There are several words exists related to misinformation. For instance, misinformation, fake news, spam, hoaxes, troll, clickbait, rumor, hate speech, etc. Different methods are used to uncover fake news. Naive Bayes (NB), Decision Trees (DT), Support Vector Machines (SVM), Linear models, Neural Networks (NN), and Ensemble models are only a few illustrations of the supervised machine learning classifiers that are employed for the same task. Techniques like Term

300

U. Chauhan et al.

Fig. 1. Fighting Misinformation on Social Media. (Image Source: https://encryptedtbn0.gstatic.com/.)

Frequency-Inverted Document Frequency (TF-IDF) and Count-Vectorizer (CV) are applied for feature extraction [2,5]. In this paper, we have proposed a novel framework for the classification of information. The framework comprises the components relevant to crowd intelligence that help the classifier to perform better. We classify Gujarati newspaper articles as fake or real using the content-based classification method. To achieve our objective, we built a dataset for Gujarati news articles. We extracted these articles from the electronic version of a newspaper. We applied several classification techniques on the dataset for separating genuine news from fake news, preceded by two feature selection methods, Count Vectorizer and Term FrequencyInverse Document Frequency (TF-IDF). In the experiment and result section, we compare the results and discuss the efficacy of the techniques applied and feature selection methods. The paper is organized as follows: Sect. 2 shows the literature review of the relevant work. It also narrates the crowd intelligence influence in misinformation detection. Then, we explain our proposed approach in Sect. 3. It is followed by the experimental result section. This section discusses corpus preparation, preprocessing phase, the classification techniques, evaluation measures, and experimental outcomes discussion. In the end, we summarize our work with conclusive remarks.

2

Literature Review

Many methods for distinguishing fake news have been invented during the last few years. We’ll talk about some of the research done to combat false information posted on social media in this part.

The Content Based Misinformation Detection for Gujarati Language

301

To classify whether the news on social media is fake or real, we need to manually label the articles as real or fake. It is a time-consuming process. Sawinder Kaur et al. For automating the detection of fake news, the Multi-level voting system has been proposed. To increase the classifier’s performance and shorten its training time for the three different datasets-news trends, Kaggle, and Reutersthey used a multi-level voting model in this technique. To address the shortcomings of the current individual models, the ML models are combined based on the least false positive (FP) rate [5]. The early identification of rumours has been the subject of numerous research projects. Zhe Zhao et al. developed a method for the early detection of rumours from the inquiry postings on social networking sites using the Gardenhose dataset and the dataset for the Boston marathon bombing. They extracted 13 statistical features of candidate clusters that were independent of any particular substantive content. After that, they trained the classifiers using these features, to obtain a better ranking function. Gardenhose dataset contains 1,242,186,946 tweets. To process such a large dataset, they used MapReduce and experimented on a 72 core Hadoop cluster (version 0.20.2) [27]. To appropriately identify the rumours, Qazvinian et al. (2011) investigated three aspects, including content-based, network-based, and microblog-specific memes. These characteristics could also be employed to try and locate users who spread false information or disinformation. The authors used 10,000 manually labeled tweets taken from Twitter for this investigation, and their Mean Average Precision (MAP) score was 0.95, according to [15]. Liang Wu et al. tried characterizing social media messages by the way they propagated from one user to other users. They suggested TraceMiner, a cuttingedge method for categorising social media communications using diffusion network data. TraceMiner receives the message’s traces as input and outputs the category [26]. TraceMiner makes use of the social network’s social dimensions [19] and the closeness of nodes-which have been effectively used to capture the unique traits of social media users in a variety of applications. Shebuti Rayana and colleagues introduced SpEagle, a novel method for the opinion spam detection problem, which integrates relational data with metadata, i.e., makes use of all available graph, behavioural, and review content data. Specifically, SpEagle uses the metadata (text, timestamps, ratings) to design and extract the characteristic features of spam, which are changed over into a spam score to be utilized as a part of class priors [17]. Gautam Kishore shahi et al. proposed a multilingual cross-domain fact check news dataset for Covid-19. The team manually annotated the dataset into 11 categories and the dataset is in 40 different languages. They had used the BERT classifier for the classification [18]. Up until this point, all research on fake news recognition has been limited to SVM and Naive Bayes classifiers, which only leverage n-gram [11] and TfIdf feature-extraction techniques. But with the advancement in deep learning, models like RNN, LSTM, etc. can be used for the classification to obtain better accuracy [7]. Autoencoders can also be used for the same.

302

U. Chauhan et al.

After reviewing existing MID research, mainly four different ways are relevant to crowd intelligence, which we have mentioned below: 1. Crowd Learning Models- To integrate crowd intelligence in MID models, it primarily uses feature engineering and representative learning [17,26,27]. 2. Crowd Behavior Modelling- To determine the credibility of a piece of information, it employs graph or probabilistic models to model crowd behavior and interactions [3,4,6,10]. 3. Crowd Knowledge Transferring- When it comes to new events, the trained MID models are normally ineffective. This method addresses the problem of transferring crowd awareness from previous events to new ones [16,24,25]. 4. Hybrid Human-Machine Models- This method focuses on creating hybrid human-machine models for MID, taking into account the complementary nature of human and machine intelligence [1,13,14,21,22].

3

Proposed Architecture

In the past few years, extensive research has been carried out in the field of fake news detection. Most of the work is done in the English language. All the datasets that have been used are in English. All researchers have used various methods to perform misinformation detection. They have trained the deep learning and machine learning models on the datasets and have obtained the results. For classification, mostly SVM and various neural networks have been used. The majority of current techniques, however, have a tendency to extract event-specific features that are hardly applicable to newly presented events [28]. According to Tolosi et al. [20], it can be challenging to detect disinformation across several events using a feature engineering-based strategy because characteristics vary significantly over time. Model generality or adaptivity is crucial for MID models to be used to a variety of events. The user’s confidence in the learning model can be increased by offering proof or explanations of the learning results. Other related disciplines, such as recommender systems, have also looked into explanations. MID explanatory models have been researched a few times. In addition to improving the acceptability, credibility, and satisfaction of recommender systems, recommender systems can also improve their persuasiveness. Fake news detection can be improved by integrating human intelligence. Typically, it is a component of the “human computation” methodology, that seamlessly merges crowd and machine skills to do tasks that neither can complete on their own [12,23]. All this work is done for the English language. We have tried to solve the above problem for regional, low-resource language using crowd wisdom or opinion from domain experts. Our work belongs to Hybrid human-machine systems. Our work shows similarity with that of Ma et al. [9]. An aggregation method based on crowd-sourcing named Fait Crowd has been proposed by him. Fait Crowd models a question’s content and the sources’ responses on a probabilistic Bayesian model to collectively learn the topic distribution and topic-based knowledge of the answer providers.

The Content Based Misinformation Detection for Gujarati Language

303

Fig. 2. Architecture for detecting misinformation with domain expert

Our proposed model tries solving the problem by taking opinions from the experts of a particular field. Considering the scenario shown in Fig. 2, our model would take opinions from the official UGC website or some experts in the field of education. As shown in Fig. 2, we have proposed the architecture for detecting misinformation based on the concept of crowd intelligence. We have collected the data from newspaper applications and newspaper websites. The collected data is then being pre-processed in such a way that the model accepts it and we can get appropriate results. In the pre-processing phase, we removed the stopwords, single-letters words, punctuation marks, symbols, digits, and duplicate words. Then we split the dataset into training and testing sets. Thereafter, the data was given to the classifier for training. To know the credibility of data, we have taken the reviews and opinions from external sources and domain experts. External sources refer to authenticated Government organizations/websites and the news channels. Domain experts refer to the people belonging to different domains such as sports, politics, and education. We believe that this type of architecture will be helpful in real-time classification of news. As shown in Fig. 3, we have explained how the domain experts would contribute to combating misinformation. The statement to be tested for credibility is given to all the domain experts. Based on the expertise, the domain experts give their opinions on the statement. For example, considering an input statement X, the domain expert 1 classifies it as 90% real, 8% take, and 2% partly true. The domain expert 2 classifies it as 10% real, 80% take, and 10% partly true. Similarly, all the domain experts give their respective opinions. Now, each domain expert is initially assigned a weight variable that specifies the accuracy of his classification. Initially, all domain experts

304

U. Chauhan et al.

Fig. 3. The role of domain experts in the misinformation detection

are assigned weights=1. Thereafter, the weighted average of the different classes is taken to make the final prediction. Once the prediction is done, it is compared with the actual news. All the domain experts who made the prediction are given some increments in their weights for high accuracy of future classifications. We have tried implementing the above approach for a single statement as input to the system. Also, we have tried comparing the statements with some official sources like government websites, newspaper sites, etc. In this approach, we have tried to find out the similarity of the input statement with the contents of the “Times of India” website. We have used the Jaccard and Cosine similarity algorithms for the same. The main challenge was to find the semantic similarity among the sentences. After comparison with the official resources, we determine the best class output using the highest similarity obtained. We believe that this approach will be more helpful in achieving more accuracy than the traditional machine learning model approach.

4

Experimental Results

4.1

Data Collection

Social media platforms such as Facebook1 , Twitter2 , etc. generate and disperse a large amount of data in various languages but majorly, the data is generated in the English language. So, to get the data in a low resource language like Gujarati, and then finding whether the news is fake or credible is a difficult and challenging task. We managed to gather the data from newspaper applications and websites 1 2

https://www.facebook.com/. https://twitter.com/.

The Content Based Misinformation Detection for Gujarati Language

305

for our experimental purpose. We have collected 242 news articles from the electronic editions of the various newspapers. The collected corpus comprised 50% of the dataset labeled as real news and the remaining 50% labeled as fake news articles. 4.2

Preprocessing

The detailed and systematic working of the experiment process can be depicted using the workflow. The input comprises a collection of newspaper articles. First, it is must to perform data preprocessing on the entire collection. Data must be preprocessed to ensure that we can extract maximum insights from the data. Datasets are highly prone to noisy, inconsistent, redundant, trivial, and incomplete data because of their huge size and more likely, their origin from various sources. Preprocessed data improves the efficiency and ease of the experiment’s purpose. When the textual document is preprocessed, non-semantic words, like prepositions, conjunctions, pronouns, etc., or stop words, are removed, since they do not contribute to retrieval of information and do not provide much information about fake content. Stopwords list differ from language to language. It is also found that there are few words in the dataset, which are not in Gujarati but Hindi or English. It is also necessary to remove these words also. The punctuation marks, digits, and symbols can also be removed. The Natural Language Tool Kit Library (NLTK) [8] is very helpful in performing these pre-processing steps. The next step is to remove the single-letter words. Although the single-letter words cannot be said to be in the category of stopwords, they don’t contribute to the text value. Rather, it is found that their removal results in better text interpretation. Hence, these are removed using the regex library. Regex library has been used in a python to define a search pattern using a sequence of characters. Redundant data can be removed using the pandas library. We have performed splitting the dataset using the split ratios by: a) 70:30 split and b) 67:33 split

4.3

Major Findings

We performed a set of experiments to examine the accuracy of various classifiers over the news article dataset. The test was performed on the Google Colab system with an Intel Xeon CPU @ 2.30 GHz * 2 processor having memory size 13.3 GB. Furthermore, we implemented all classifier methods on Python version 3.5. The corpus comprised 242 news articles, which were splitted into 70:30 ratio for training and testing. The major findings deal with the point that the classifier built using the training dataset could give accurate and efficient results for the testing dataset. It can be inferred that the ability of the classifier to distinguish the fake and the real content depends upon factors such as the corpus and the classification model used. Table 1 comprises the accuracy of the classifiers for both the feature selection measures, TF-IDF and CV. As provided in Table 1, 8 of 13 models

306

U. Chauhan et al.

achieved an accuracy more than 80%, and for the rest, the accuracy falls below 80%. From the Precision perspective, Stochastic Gradient Descent (88%) performs best using TF-IDF, and Nu SVC (88%) perform best using CV for the corpora. From the Recall perspective, Stochastic Gradient Descent(88%) performs best using TF-IDF, and Nu SVC (88%) performs best using CV for the corpora. A classifier can be considered appropriate if and only if it achieves a high precision as well as high recall. Hence, to obtain a generalized result, F1score is calculated. Upon evaluating the F1-score, it can be again observed that Stochastic Gradient Descent(88%) performs best using TF-IDF, and Nu SVC (88%) perform best using CV. Table 1. Comparative Analysis using Machine Learning Classifiers Model no. Models

Accuracy (%)

Precision (%) Recall (%)

TFIDF CV

TFIDF CV

F1-score (%)

TFIDF CV TFIDF CV

A1

Multinomial Na¨ıve Bayes

76.71

79.45 78

81

77

80

76

79

A2

Gaussian Niave Bayes

73.97

72.6

75

74

73

73

72

A3

Complement Na¨ıve Bayes

78.08

79.45 79

81

78

80

78

79

A4

Bernoulli Na¨ıve Bayes

78.23

77.12 83

83

78

78

77

77

A5

Stochastic Gradient Descent

87.67

83.56 88

84

88

84

88

84

A6

Logistic Regression

84.93

86.3

86

87

85

86

85

86

A7

Perceptron

84.93

80.82 85

81

85

81

85

81

A8

Random Forest

80.82

84.93 81

86

81

85

81

85

A9

Linear SVC

84.93

86.3

87

85

86

85

86

A10

Multi-Layer Perceptron

86.3

83.56 87

84

86

84

86

84

A11

Nu SVC

82.19

87.67 83

88

82

88

82

88

A12

Decision Tree

78.08

71.23 79

71

78

71

78

71

A13

Adaboost

84.93

82.19 85

83

85

82

85

82

76

86

Besides, we compared the accuracy of all models with respect to TF-IDF and CV. Figure 4 shows the bar chart accuracy discovered on the training and testing division by splitting the corpus in 70:30. The outcomes revealed that TF-IDF dominates the accuracy column over the CV up to some extent. Furthermore, we performed experiments for top 3 classifiers, having TF-IDF as feature selection, to assess the outcomes of precision versus recall value. Figure 5 shows the precision value when the recall value reaches its peak. It achieved a value between 0.5 and 0.6 for the highest recall value. It can be observed that the recall values fall and rise by a negligible magnitude with a decrease in the precision values. As shown in Fig. 6, similar effect has been found while testing with the top 3 accurate classifiers having CV as a feature selection. In this experimental setup too, we discovered that recall values find a small deviation with respect to decreasing the precision values. The SGD classifier sees more deviating values of recall compared to the other two methods, Logistic Regression and Nu SVC having step-wise manner fall in the precison.

The Content Based Misinformation Detection for Gujarati Language

307

Fig. 4. Accuracy

(a) NuSvc

(b) Gradient Descent

(c) LinearSvc

Fig. 5. Top 3 TFIDF

(a) Logistic Regression

(b) NuSvc

(c) Gradient Descent

Fig. 6. Top 3 CV

5

Conclusion

In this paper, we proposed the architecture for classifying the news as fake or credible. As fake news or misinformation carries some special characteristics, we identified them and exploited to prevent mobile users or internet users

308

U. Chauhan et al.

from spreading fake information. We showed how the domain experts, fact-check source and crowed intelligence components can be integrated in the misinformation detection approach. We implemented the several models in order to examine the efficacy of those over the misinformation detection domain, we found out that the model performed better those uses the fact check platform. We could observe that neural-based techniques performed better than other techniques such as Naive Bayes and decision tree induction.

References 1. Bhattacharjee, S.D., Talukder, A., Balantrapu, B.V.: Active learning based news veracity detection with feature weighting and deep-shallow fusion. In: 2017 IEEE International Conference on Big Data (Big Data), pp. 556–565. IEEE (2017) 2. Guo, B., Ding, Y., Yao, L., Liang, Y., Yu, Z.: The future of misinformation detection: new perspectives and trends (2019). arXiv:1909.03654 3. Hooi, B., Shah, N., Beutel, A., G¨ unnemann, S., Akoglu, L., Kumar, M., Makhija, D., Faloutsos, C.: Birdnest: Bayesian inference for ratings-fraud detection. In: Proceedings of the 2016 SIAM International Conference on Data Mining, pp. 495–503. SIAM (2016) 4. Jin, Z., Cao, J., Zhang, Y., Luo, J.: News verification by exploiting conflicting social viewpoints in microblogs. In: Thirtieth AAAI Conference on Artificial Intelligence (2016) 5. Kaur, S., Kumar, P., Kumaraguru, P.: Automating fake news detection system using multi-level voting model. Soft Comput. 24(12), 9049–9069 (2019). https:// doi.org/10.1007/s00500-019-04436-y 6. Kumar, S., Hooi, B., Makhija, D., Kumar, M., Faloutsos, C., Subrahmanian, V.: Rev2: fraudulent user prediction in rating platforms. In: Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining, pp. 333–341 (2018) 7. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521(7553), 436–444 (2015) 8. Loper, E., Bird, S.: Nltk: the natural language toolkit. cs/0205028 (2002) 9. Ma, F., Li, Y., Li, Q., Qiu, M., Gao, J., Zhi, S., Su, L., Zhao, B., Ji, H., Han, J.: Faitcrowd: fine grained truth discovery for crowdsourced data aggregation. In: Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 745–754 (2015) 10. Ma, J., Gao, W., Wong, K.F.: Rumor detection on twitter with tree-structured recursive neural networks. Association for Computational Linguistics (2018) 11. Magdy, A., Wanas, N.: Web-based statistical fact checking of textual documents. In: Proceedings of the 2nd International Workshop on Search and Mining Usergenerated Contents, pp. 103–110 (2010) 12. Michelucci, P., Dickinson, J.L.: The power of crowds. Science 351(6268), 32–33 (2016) 13. Nguyen, A.T., Kharosekar, A., Krishnan, S., Krishnan, S., Tate, E., Wallace, B.C., Lease, M.: Believe it or not: designing a human-ai partnership for mixed-initiative fact-checking. In: Proceedings of the 31st Annual ACM Symposium on User Interface Software and Technology, pp. 189–199 (2018)

The Content Based Misinformation Detection for Gujarati Language

309

14. Nguyen, A.T., Kharosekar, A., Lease, M., Wallace, B.: An interpretable joint graphical model for fact-checking from crowds. In: Thirty-Second AAAI Conference on Artificial Intelligence (2018) 15. Qazvinian, V., Rosengren, E., Radev, D., Mei, Q.: Rumor has it: Identifying misinformation in microblogs. In: Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pp. 1589–1599 (2011) 16. Qian, F., Gong, C., Sharma, K., Liu, Y.: Neural user response generator: fake news detection with collective user intelligence. In: IJCAI, vol. 18, pp. 3834–3840 (2018) 17. Rayana, S., Akoglu, L.: Collective opinion spam detection: bridging review networks and metadata. In: Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 985–994 (2015) 18. Shahi, G.K., Nandini, D.: Fakecovid–a multilingual cross-domain fact check news dataset for COVID-19 (2020). arXiv:2006.11343 19. Tang, L., Liu, H.: Relational learning via latent social dimensions. In: Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 817–826 (2009) 20. Tolosi, L., Tagarev, A., Georgiev, G.: An analysis of event-agnostic features for rumour classification in twitter. In: Tenth International AAAI Conference on Web and Social Media (2016) 21. Tschiatschek, S., Singla, A., Gomez Rodriguez, M., Merchant, A., Krause, A.: Fake news detection in social networks via crowd signals. In: Companion Proceedings of the Web Conference 2018, pp. 517–524 (2018) 22. Vo, N., Lee, K.: The rise of guardians: fact-checking url recommendation to combat fake news. In: The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, pp. 275–284 (2018) 23. Von Ahn, L.: Human computation. In: 2008 IEEE 24th International Conference on Data Engineering, pp. 1–2. IEEE (2008) 24. Wang, Y., Ma, F., Jin, Z., Yuan, Y., Xun, G., Jha, K., Su, L., Gao, J.: Eann: event adversarial neural networks for multi-modal fake news detection. In: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 849–857 (2018) 25. Wu, L., Li, J., Hu, X., Liu, H.: Gleaning wisdom from the past: early detection of emerging rumors in social media. In: Proceedings of the 2017 SIAM International Conference on Data Mining, pp. 99–107. SIAM (2017) 26. Wu, L., Liu, H.: Tracing fake-news footprints: characterizing social media messages by how they propagate. In: Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining, pp. 637–645 (2018) 27. Zhao, Z., Resnick, P., Mei, Q.: Enquiring minds: Early detection of rumors in social media from enquiry posts. In: Proceedings of the 24th International Conference on World Wide Web, pp. 1395–1405 (2015) 28. Zubiaga, A., Aker, A., Bontcheva, K., Liakata, M., Procter, R.: Detection and resolution of rumours in social media: A survey. ACM Comput. Surv. (CSUR) 51(2), 1–36 (2018)

Analysis of Indicators in the Urban Distribution of Food in Megacities Within the Framework of Sustainability Leila Nayibe Ram´ırez Casta˜ neda1,2(B) , Sonia Lucila Meneses1,2 , Astrid Altamar1,2 , and Edwin Bulla1,2 1

2

Universidad Libre, Bogot´ a D.C., Colombia {leylan.ramirezc,sonial.menesesv, astridd.altamarc,edwina.bullap}@unilibre.edu.co Facultad de Ingenier´ıa, Programa de Ingenier´ıa Industrial, Bogot´ a D.C., Colombia

Abstract. The analysis of the sustainability of urban food distribution systems in cities is a complex task that involves significant technological efforts, as well as the ability to analyze the available information of the indicators to generate an aggregate measure to understand and understand the performance of supply chains in urban centers. The information for the quantification of these indicators suggested from the literature is scarce and its processing adds value to those interested in establishing holistic solutions that benefit stakeholders from the public, private companies and society to achieve efficient solutions with the least negative impact on load distribution. This article allows us to study some tools of aggregate indicators to measure the sustainability of distribution systems in megacities and from these results propose smart urban logistics solutions for supply chains that supply millions of people in megacities.

Keywords: Urban distribution of goods sustainability analysis

1

· multi-criteria evaluation ·

Introduction

The goal of Urban Logistics (UL) is to reconcile the interests of all stakeholders and find compromise solutions to problems associated with urban logistics flows [14]. However, this field of research increasingly appropriates innovative strategies to consider alternatives that contribute to the distribution of goods within the city under the framework of sustainability to build cities that allow a balance between the economic growth of flexible transport and social promotion that reduces the negative environmental effects of the externalities generated by the Supported by Organization Universidad Libre. c The Author(s), under exclusive license to Springer Nature Switzerland AG 2023  L. F. Castillo Ossa et al. (Eds.): SSCT 2023, LNNS 732, pp. 310–318, 2023. https://doi.org/10.1007/978-3-031-36957-5_27

Analysis of Indicators in the Urban Distribution of Food in Megacities

311

flow of goods [6]. Priority dimensions of sustainability include the economic perspective (e.g. economic returns, markets), environmental (e.g. emissions, water), and social aspects (e.g. occupational health and safety) [5], indices of concern to urban logistics that contribute significantly to the sustainability of megacities around the world [14]. Also, Sustainable Development Goal 11 recommend ensuring sustainable transport systems, as well as the entire urban ecosystem [1]. The concept of a smart city has increased its popularity in the field of research since sustainability is a governance framework that motivates the construction of cities that prioritize a prospective towards intelligent mobility of all modes of transport available in the city as well as for multiple purposes, but to achieve this desired scenario requires information technologies, real-time processing, and analysis to incorporate decision-making that connects the government with the needs of citizens, within the available technologies is the Internet of Things (IoT), developments that in the future will reduce the negative environmental impacts of the movements of goods, and determine restrictive policies that allow their viability from social and economic aspects [2]. On the other hand, regarding management in the urban distribution of goods, public authorities usually propose solutions from investment in public infrastructure, which implies a reorganization of the logistics flow in cities but sometimes this implies a reduction in efficiency, reducing the margins of economic contribution to distribution operations, the alternatives must be reasonable and the support of their viability can be possible from the implementation of monitoring technology, which allows detailing the cost-benefit analysis of their implementation [10]. The purpose of the study was to obtain information from secondary sources of information for logistics measures of the city from different dimensions of sustainability. The authors try to verify the hypothesis: There is a significant relationship between the different selected economic, social and environmental variables that affect the urban distribution of goods in the most important road corridor of the city and sustainability. The article is structured as follows: we present the concept of logistics city and environmental sustainability clearly to provide the framework in the introduction. Section 2 describes the methodology used. The next section presented the analysis and data results in the paper. The article ends with the conclusions and future research.

2

Methodology

For this phase, the proposed methodology is considered [11]. The data generated by the implementation of sensors in smart cities have allowed solutions to problems with better solutions, this condition implies the need to implement big data analytic techniques in terms of algorithms, types of data and, tools for their analysis. Big Data requires technologies for capturing, storing, managing and analyzing a high volume of information that creates highly complex tasks, due to its variety and size [4].

312

L. N. Ram´ırez Casta˜ neda et al.

The methodological and structural scope was established with information from the past available from the selected indicators with categories related to sustainability and the structure is proposed to achieve the complete development of a Big Data Analytic system for the Urban Distribution of the food sector that transits through the road corridor of 13th Street. It was found with a limited and updated availability of information from different sources managed by public agencies in the city between 2019 and 2021, establish conditions of correlation analysis, descriptive statistics and inferential statistics, with data from the past to validate them with real-time data that allows knowing the deviation of predictive and prescriptive algorithms compared to the real information that can be reviewed in a next phase of the project. Some recommended implementation software for the technological development of these applications is Hadoop, spark, tableu, google trends, Mobile Edge Computing Framework, HPCC, stratosphere, IBM, Rapid Miner, R, Apache Mahout, NoSQL, MDM [11]. The interest of this research is focused on the development of open access applications such as R and Python. In relation to the stages for the development of this research [8], the initial phase consisted of preparing the analysis data, the preprocessing of the data and its transformation was carried out using specific RStudio libraries (2023). The data transformation phase involved the technique of least squares scaling and normalization [3], defined by RStudio as The root-mean-square for a (possibly centered) column is defined as where x is a vector of the non-missing values and n is the number of non-missing values. In the case center = TRUE, this is the same as the standard deviation, but in general it is not. (To scale by the standard deviations without centering, use scale(x, center = FALSE, scale = apply(x, 2, sd, na.rm = TRUE)) [13]. Followed by the treatment of the information, an output variable was constructed that is the weighted sum of the indicators selected for the object of study, which has been defined as sustainability, the selection of variables under the criteria of completeness and availability allows to improve the precision, the statistical assumptions are verified to establish correlations with parametric and nonparametric test according to the nature of the variable. The normality tests used was the Shapiro wilk test [12]. Correlation analyses were applied to each of the variables with respect to the output variable. In relation to the calculation of CO2 emissions, the equation proposed by the European Association for forwarding, Transport, Logistics and Customs Services (CLECAT) [9] is considered. According to this methodology, greenhouse gases are obtained with the equations [8]: GT = Tcap × gt

(1)

(Wg − ϕ  χ) × d (2) 1000 where GT stands for total greenhouse gas emissions per tonne kilometres (g CO2 e/t.km), Tcap for transport capacity, gt for emissions factor which depends on the vehicle weighting (larger vehicles usually have a smaller emission factor), Tcap =

Analysis of Indicators in the Urban Distribution of Food in Megacities

313

Wg for gross weight (laden weight), and d for distance. Factors for the calculation of energy consumption and greenhouse gas emissions (calculated as CO2 equivalents) in accordance with EN 16258 [9]. 2.1

Data Selected for Analysis

The following table, based on the previous bibliographic review of indicators for the management of Urban Distribution Goods, this first approximation may vary according to the availability of information and validation by experts thus, the analyzed time horizon of the information corresponds to July 2019 to June 2021 (see Table 1): Table 1. Multivariate model coefficient Objective

Indicators

Category

Source of information

Define strategies Cargo vehicle that allow the safe violations circulation of cargo flow in Bogot´ aRegion Accidents in cargo vehicles

Social

SDM

Establish CO2 emissions from environmental freight transport impacts, for the (Million Tons/year) generation of policies, regulations or incentive programs Air quality level in identified logistics zones (particulate matter) Compliance with noise regulations in the industrial sector

Environmental SDA

Achieve supply chain efficiencies in the most impactful economic sectors Reducing logistics costs has an impact on the final price of the product

Economic

Load transport speed (km/hr)

Share of Bogot´ a’s logistics costs in the price of the product per supply chain

SDM

SDA

SDA

SDM

Ministry of Transport

314

2.2

L. N. Ram´ırez Casta˜ neda et al.

Sustainability Indicator Model

The model proposed for this study allows to show the relationship between the independent variables (selected indicators) and the independent variable (Sustainability indicator). This in order to allow us to understand the behavior of the variables in the face of sustainability and determine if this multiple regression model [7]allows to explain sustainability through the collection of data considered as independent variables within the model, in order to give interpretation to the phenomenon studied in this article. The objective is to determine the weights of the independent variables with respect to the independent variable using multiple regression as a descriptive tool. Y = βX + ε

(3)

In the above equation Ynxq is the matrix of observations of the response variables, with each row representing a case (or observation set) and each column representing a characteristic of the response. The matrix Xnxk is the observed matrix of the corresponding explanatory variables (as in the univariate response linear model considered before) [7].

3 3.1

Results Descriptive Analysis of Variables

According to Fig. 1, It can be observed each of the variables analyzed for the study of the aggregate indicator of sustainability through the information available in the time horizon defined in this analysis it can be observed that the variables from the shapiro wilks test present a normal behavior with a confidence

Fig. 1. Histograms for selected variables

Analysis of Indicators in the Urban Distribution of Food in Megacities

315

level of 0.05 are noise level measured through the LRAeq unit, the number of cargo accidents, PM10 particulate matter emissions, average speed, total trips, kilograms of cargo, gallons of fuel, and sustainability added value. On the other hand, the variables that do not show this behavior were CO2 emissions, kilometers traveled and the values paid to enter food through the 13th Street road corridor. 3.2

Multivariate Model

Table 2 presents the coefficients corresponding to the model of the aggregate indicator of sustainability that can be explained through a multiple regression model, with adjusted R of 1. According to the results obtained, the independent variables explain the aggregate indicator of sustainability since the significance is less than 0.05. Table 2. Multivariate model coefficient Estimate Std. error t value

P r (> |t|)

(Intercept)

2.94E–17 1.78E–17 1.65E+00 0.129

LRAeq day

7.69E–02 3.04E–17 2.53E+15