Pan-African Artificial Intelligence and Smart Systems: Second EAI International Conference, PAAISS 2022 Dakar, Senegal, November 2–4, 2022 Proceedings

Preface
Organization
Contents
IoT and Enabling Smart System Technologies
A Certificate-Based Pairwise Key Establishment Protocol for IoT Resource-Constrained Devices
	1 Introduction
	2 Related Works
	3 Vocabulary and System Model
		3.1 Design Parameters and Vocabulary
		3.2 Design Goal
		3.3 Proposed Key Establishment Scheme
		3.4 Setting Up the Certificates and the Symmetric Key
	4 Proof of Security
	5 Performance Evaluation
		5.1 Simulation Parameters
		5.2 Results
		5.3 Scalability Analysis
		5.4 Security Analysis
	6 Comparison with Related Work
	7 Conclusion
	References
Reinforcement Learning-Based Dynamic Path Allocation in IoT Systems
	1 Introduction
		1.1 Limitations and Challenges
		1.2 Authors' Contributions
	2 Related Work
	3 System Model
		3.1 Network Assumptions
		3.2 Graph Theory Representation of IoT Network
		3.3 Design Goal and Problem Formulation
		3.4 Terminologies
		3.5 Reinforcement Learning Framework
	4 Simulation
		4.1 Network Topology and Parameters of Simulation
		4.2 Quantitative Evaluation
		4.3 Statistical Data Analysis
	5 Conclusion
	References
Reduction of Data Transmission in an IoT Wireless Sensor Network
	1 Introduction
		1.1 Contributions
	2 Related Work
	3 Methodology
		3.1 Data Transmission Problem Formulation
		3.2 k-Nearest Neighbours (k-NN)
		3.3 Recurrent Neural Network (RNN)
		3.4 Energy System
		3.5 Overview of the Method
		3.6 Test Data Set
	4 Performance Evaluation
		4.1 Model Training
		4.2 Node Selection
		4.3 Energy Consumption Analysis
	5 Conclusion
	References
Special Topics of African Interest
A Comparative Study of Regressors and Stacked Ensemble Model for Daily Temperature Forecasting: A Case Study of Senegal
	1 Introduction
		1.1 Projected Climate Changes
		1.2 Brief Review of Similar Studies
		1.3 Study Location and Source of Data
		1.4 Exploratory Data Analysis
	2 Materials and Methods
		2.1 Machine Learning Models
		2.2 Dataset Restructuring
		2.3 Model Evaluation Metrics
	3 Results and Discussion
		3.1 Maximum Temperature
		3.2 Minimum Temperature
		3.3 Forecasts on the Test Dataset
		3.4 Conclusion
	4 Limitation of the Study
	References
Exploring Use of Machine Learning Regressors for Daily Rainfall Prediction in the Sahel Region: A Case Study of Matam, Senegal
	1 Introduction
		1.1 Effects of Climate Change on Rainfall in Senegal
		1.2 Weather Forecasting
		1.3 Weather Prediction Using Machine Learning
		1.4 Significance and Objective of the Study
		1.5 Study Area Geographical Characteristics and Data Source
		1.6 Exploratory Data Analysis for Rainfall
	2 Review of Related Work
	3 Materials and Methods
		3.1 Data Pre-processing, Transformation and Feature Engineering
		3.2 Correlation of Features
		3.3 Predictors and Splitting of the Dataset
		3.4 Machine Learning Models
		3.5 Model Configuration and Performance Improvement
		3.6 Model Evaluation
	4 Results and Discussion
		4.1 Gradient Boosting Regressor Results
		4.2 Random Forest Regressor Results
		4.3 CatBoost Regressor Results
		4.4 Ridge Regression Results
		4.5 Forecasting on the Test Dataset
	5 Conclusion
	6 Our Future Work
	References
Artificial Intelligence Theory and Methods
Dynamic Pre-trained Models Layer Selection Using Filter-Weights Cosine Similarity
	1 Introduction
		1.1 Contributions
	2 Literature Review
	3 Methodology
		3.1 Cosine Similarity
		3.2 Proposed Approach
		3.3 Experiments
	4 Results and Discussion
		4.1 Comparison Between the Selected Methods
		4.2 Comparison of Selection Methods and the Baselines
	5 Conclusion
	References
Learning Approximate Invariance Requires Far Fewer Data
	1 Introduction
	2 Related Works
		2.1 Data Augmentation Research
		2.2 Posterior Training
		2.3 Encoding Prior Knowledge
	3 Methodology
		3.1 Online and Offline Data Augmentation
		3.2 Bayesian Neural Networks
		3.3 Learning Approximate Invariant Priors
		3.4 A Lightweight Formulation
	4 Experiments
		4.1 Sample Size and Augmentation Factor
		4.2 Stability of Invariance Inducing Techniques on Generalization Gains
		4.3 Effect of Invariance Inducing Technique on Uncertainty
	5 Conclusion
	References
Deep Matrix Factorization for Multi-view Clustering Using Density-Based Preprocessing
	1 Introduction
		1.1 Overview
		1.2 Problems Introduction
	2 Proposed Approach
		2.1 Density-Based Data Preprocessing
		2.2 Semi-Non-Negative Matrix Factorization
		2.3 Optimization Algorithm
	3 Experimentation
		3.1 Datasets
		3.2 Benchmarks
	4 Results and Analysis
		4.1 Results Using Yale Dataset
		4.2 Results using Extended YaleB Dataset
		4.3 Covergence Analysis
	5 Conclusion
	References
Application of Genetic Algorithm for Complexity Metrics-Based Classification of Ontologies with ELECTRE Tri
	1 Introduction
	2 Related Work
	3 Materials and Methods
		3.1 Ontology Quality Dimensions
		3.2 ELECTRE Tri
		3.3 Threshold Inference with the Genetic Algorithm
	4 Experimental Results and Discussion
		4.1 Dataset
		4.2 Computer and Software Environment
		4.3 Class Definition
		4.4 Ontology Classification
	5 Conclusion
	References
Local Features Based Spectral Clustering for Defect Detection
	1 Introduction
	2 Materials and Methods
		2.1 Local Binary Patterns
		2.2 Grey Level Co-occurrence Matrix (GLCM)
		2.3 Spectral Clustering
	3 Results and Discussion
		3.1 Combination LBP and Spectral Clustering
		3.2 Combination GLCM and Spectral Clustering
	4 Conclusion
	References
Artificial Intelligence Applications in Medicine
Help in the Early Diagnosis of Liver Cirrhosis Using a Learning Transfer Method
	1 Introduction
	2 Methodologies
		2.1 Presentation of the Data
		2.2 Deep Learning and Transfer Learning in Liver Image Classification
		2.3 Statistical Analysis
	3 Results
		3.1 Results of the Classification
		3.2 Confusion Matrix
		3.3 ROC Curve
	4 Discussion
		4.1 Discussion Following the Results of Other Models (VGG16, AlexNet, GoogleNet)
		4.2 Discussion Following the Results of the Literature Articles
	5 Conclusion
	References
Automatic Detection of COVID-19 Using Ensemble Transfer Learning Based on Lung CT Scans
	1 Introduction
	2 Related Works
	3 Methods and Techniques
		3.1 Dataset
		3.2 Image Pre-processing
		3.3 Training Process
		3.4 Transfer Learning
		3.5 Convolutional Neural Network Selection
		3.6 Ensemble
		3.7 Hardware Specifications
	4 Results and Discussions
		4.1 Evaluation Metrics
		4.2 Discussions
		4.3 Comparisons with Related Works
	5 Conclusion
	References
Pancreas Instance Segmentation Using Deep Learning Techniques
	1 Introduction
		1.1 Semantic Segmentation
		1.2 Instance Segmentation
	2 UNet
		2.1 Encoder
		2.2 Decoder
	3 Proposed Model
		3.1 Residual Networks (ResNet)
		3.2 Watershed
	4 Methods and Techniques
		4.1 Max Pooling Operation
		4.2 ResNet-34 Encoder
	5 Experiments
		5.1 Generate Border Classes Steps
		5.2 Implementing UNet Plus Watershed
		5.3 TensorFlow Implementation of UNet
		5.4 Training
	6 Results
	7 Conclusion
	References
Automating Sickle Cell Counting Using Object Detection Techniques
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Yolov5
		3.2 Loss Function
		3.3 Grid Sensitivity
	4 Experiments and Results
		4.1 Datasets
		4.2 Training
		4.3 Results
		4.4 Deployment
	5 Conclusion and Future Work
	References
Remote sensing and AI in Agriculture
Quaternionic Wavelets for Estimating Forest Biomass by Gradient Boosting Methods
	1 Introduction
	2 Aboveground Biomass Estimation
		2.1 FOTO Method
		2.2 FOTO+ Method
	3 Using Quaternionic Wavelets Transform for Descriptors Extraction
		3.1 Classic Wavelet Transform
		3.2 Quaternionic Wavelet Transform (QWT)
		3.3 Using Wavelets
	4 Extreme Gradient Boosting Methods for AGB Estimation
	5 Implementation
		5.1 Test Data
		5.2 Implementation Strategies
		5.3 Descriptor Extraction
	6 Results and Discussion
		6.1 Results
		6.2 Discussions
	7 Conclusion and Perspectives
	References
Detection and Classification of Underwater Acoustic Events
	1 Introduction
	2 Materials and Methods
		2.1 Methodology
		2.2 Models and Metrics
		2.3 Sensitivity Tests
		2.4 Application to Large Data Set
	3 Results
		3.1 Parameters Identification and Sensitivity Tests
		3.2 Sound Event Time Series
	4 Discussion
	5 Conclusion
	References
Plant Diseases Detection and Classification Using Deep Transfer Learning
	1 Introduction
	2 Review of Existing Work
		2.1 Artificial Intelligence (AI)
		2.2 Machine Learning (ML)
		2.3 Deep Learning (DL)
		2.4 Transfer Learning (TL)
		2.5 Convolutional Neural Network (CNN)
	3 Methods and Techniques
		3.1 Preprocessing
		3.2 Image Augmentation
		3.3 Image Rotation
		3.4 Image Shift
		3.5 Image Blurring
		3.6 Data Segmentation
		3.7 Result and Discussion
		3.8 Resnet50 Training
		3.9 Resnet50 Model Prediction
	4 Results and Discussion
	5 Conclusion
		5.1 Contributions
	References
Light-Weight Deep Learning Framework for Automated Remote Sensing Images Classification
	1 Introduction
	2 Related Works
		2.1 Research Approach
	3 Methods and Techniques
		3.1 Methodology Overview
		3.2 Proposed Model
	4 Experimental Results
		4.1 Dataset
		4.2 Results and Discussion
	5 Conclusion
	References
AI applications and Smart Systems Technologies
DeepMalOb: Deep Detection of Obfuscated Android Malware
	1 Introduction
	2 Related Work
	3 Proposal Approach to Detection of Malware Obfuscated
		3.1 Proposed Approach to the Detection of Obfuscated Malware
		3.2 Supervised Neural Network Model
	4 Experiment Setup
		4.1 Data Set Description
		4.2 Experience Environment
		4.3 Evaluation Metrics
	5 Results and Analysis
		5.1 Choice of Hyper-Parameters
		5.2 Interpretation of Results
	6 Discussion
	7 Conclusion
	References
Reconfigurable Intelligent Surfaces: Redefining the Entirety of Wireless Communication Systems in Leaps and Bounds
	1 Introduction
	2 Motivation
	3 Reconfigurable Intelligent Surface
		3.1 RIS Operation
		3.2 RIS Features
		3.3 RIS Applications
		3.4 RIS Outlook
		3.5 RIS Capacity and Spectral Efficiency
	4 Conclusion
	References
A New Class of DC-Free Run-Length Limited Codes
	1 Introduction
	2 Background
		2.1 RLL Codes in VLC Channel
		2.2 Error Correction Analysis of RLL Codes
	3 Improved RLL Code
	4 Analysis and Discussions
		4.1 Redundancy Study
		4.2 Error Correction Performance
	5 Conclusion
	References
Affective Computing
Conversational Pattern Mining Using Motif Detection
	1 Introduction
	2 Background
		2.1 Conversational Mining
	3 Methodology
		3.1 Sequence Creation Process
		3.2 Motif Detection
	4 Results
		4.1 Motif Detection Results
		4.2 Application
	5 Conclusion
	References
Facial Expression Recognition with Manifold Learning and Graph Convolutional Network
	1 Introduction
	2 Related Works
	3 Manifold Graph Convolutional Network Model
		3.1 Isomap Manifold
		3.2 Graph Convolution Network
	4 Experiment
		4.1 Datasets
		4.2 Data Preprocessing
		4.3 Experimental Setup
	5 Experimental Results and Discussion
	6 Conclusion
	References
Speech Emotion Classification: A Survey of the State-of-the-Art
	1 Introduction
		1.1 Emotion Classification
	2 Speech Emotion Classification
		2.1 Speech Emotion Classifiers
		2.2 Support Vector Machine (SVM)
		2.3 Hidden Markov Model (HMM)
		2.4 K-Nearest Neighbor (KNN)
		2.5 Decision Tree (DT)
	3 Performance Evaluation Metrics for the State-of-the-Arts Classification Algorithms
	4 Speech Emotion Classification: A Critical Analysis of State-of-the-Art Techniques and Algorithms
	5 Conclusion
	References
Intelligent Transportation Systems
Forward Obstacle Detection by Unmanned Aerial Vehicles
	1 Introduction
	2 Related Work
	3 Methodology
	4 Fast Object Detection
	5 Proposed Approach
		5.1 Presentation of the Function avoid_zone()
		5.2 Description of free_directions() and best_direction() Functions
	6 Experimental Results and Discussion
		6.1 Hardware Specifications
		6.2 Model and Processing
		6.3 Test on Raspberry Pi 4b
		6.4 Discussion
	7 Conclusion and Future Work
	References
Multi-agent Reinforcement Learning Based Approach for Vehicle Routing Problem
	1 Introduction
	2 Related Work
	3 System Model
	4 Evaluation
	5 Conclusion
	References
An On-Site Collaborative Approach of Road Crash Data Collection
	1 Introduction
	2 Background
	3 Related Work
	4 The Collaborative Collection System
		4.1 Repositories
		4.2 Sharing Crash Identifier
		4.3 Conflicts Resolution
		4.4 Security
	5 Testing of the Collaborative Collection
		5.1 Results
	6 Conclusion and Future Work
	References
Author Index