Soft Computing in Data Science: 7th International Conference, SCDS 2023, Virtual Event, January 24–25, 2023, Proceedings

Description

This book constitutes the refereed proceedings of the 7th International Conference on Soft Computing in Data Science, SCDS 2023, which was held virtually in January 2023.  The 21 revised full papers presented were carefully reviewed and selected from 61 submissions. The papers are organized in topical sections on ​​​artificial intelligence techniques and applications; computing and optimization; data analytics and technologies; data mining and image processing; mathematical and statistical learning. Preface Organization Contents Artificial Intelligence Techniques and Applications Explainability for Clustering Models 1 Introduction 1.1 Explainable AI 1.2 Limitations of Explainability AI 1.3 Probabilistic Clustering 1.4 Unsupervised Explainability 2 Related Works 3 Methodology 3.1 Supervised Explanations 3.2 Proposed Unsupervised Explainable AI (UXAI) Method 3.3 Explanations 3.4 Validation Process 4 Experiments 4.1 Datasets 4.2 Validating UXAI 4.3 Observations and Results 5 Conclusion References Fault Diagnosis Methods of Deep Convolutional Dynamic Adversarial Networks 1 Introduction 2 Transfer Learning 3 Deep Convolutional Dynamic Adversarial Networks 3.1 Feature Extraction 3.2 Dynamic Adversarial Learning Strategies 3.3 Loss Function 3.4 Proposed Method Steps and Flow Chart 4 Test Verification 4.1 Source of Data 4.2 Comparative Analysis of Experiments 4.3 Analysis of Test Results 5 Conclusion References Carbon-Energy Composite Flow for Transferred Multi-searcher Q-Learning Algorithm with Reactive Power Optimization 1 Introduction 2 Reactive Power Optimization Model of Carbon Energy Composite Flow 2.1 Basis of Carbon-Energy Composite Flow Model 2.2 Reactive Power Optimization Model Considering Carbon Energy Composite Flow 3 Migrating Multi-searcher Q Learning Algorithms 3.1 Information Matrix State-Action 3.2 Space Dimensionality Reduction 3.3 Combined Prediction Model 4 Online Transfer Learning 5 Conclusion References Multi-source Heterogeneous Data Fusion Algorithm Based on Federated Learning 1 Introduction 2 Related Work 2.1 Improved Federated Weighted Average Algorithm 2.2 Multi-source Heterogeneous Data Fusion 3 First Section Multi-source Heterogeneous Data Fusion Based on Federated Learning 3.1 Improved Federal Weighted Average Algorithm 3.2 Overall Design of the Algorithm 3.3 Sub-module Design 4 Experimental Results and Analysis 4.1 Experimental Analysis of Single-Node Heterogeneous Data Fusion 4.2 Experimental Analysis of Multi-node Heterogeneous Data Fusion 5 Conclusion References Dynamic Micro-cluster-Based Streaming Data Clustering Method for Anomaly Detection 1 Introduction 2 Related Work 3 Methodology 3.1 Cluster Structure 3.2 Algorithm of DMADSD 4 Experiment Findings and Analysis 4.1 Dataset 4.2 Detection Performance Evaluation 4.3 Computational Complexity Evaluation 5 Conclusion References Federated Ensemble Algorithm Based on Deep Neural Network 1 Introduction 2 Basic Knowledge 2.1 Deep Learning 2.2 Federated Learning 2.3 Integrated Learning 2.4 Algorithm Description 3 Implementation of Federated Integration Algorithm Based on Deep Neural Network 3.1 The Basic Idea of the Algorithm 3.2 Description of the Algorithm 3.3 Performance Analysis 3.4 Complexity Analysis of Algorithms 3.5 Security Analysis of Algorithms 4 Experimental Analysis 4.1 Experimental Setup 4.2 Experimental Analysis 4.3 Experimental Summary 5 Conclusion References Performance Comparison of Feature Selection Methods for Prediction in Medical Data 1 Introduction 2 Related Work 3 Material and Methodology 3.1 Data 3.2 Feature Selection Methods 3.3 Summary of Feature Selection Methods 4 Result and Discussion 4.1 Feature Selection Comparison’s Result 4.2 Discussion 5 Conclusion References An Improved Mask R-CNN Algorithm for High Object Detection Speed and Accuracy 1 Introduction 2 Research on MASK-R-CNN Algorithm 2.1 MASK-R-CNN Principle 2.2 Improved MASK R-CNN Algorithm 3 Experiment and Result Analysis 4 Summary References Computing and Optimization Federated Learning with Class Balanced Loss Optimized by Implicit Stochastic Gradient Descent 1 Introduction 2 Client-Server Federated Learning Update Architecture 3 Design of Federated Learning Algorithm for Implicit Stochastic Gradient Descent Optimization 3.1 Federation Nearest Neighbor Optimization 3.2 Global Model Update Optimization Based on Implicit Stochastic Gradient Descent 3.3 Types of Balance Loss 4 Experiments and Results 4.1 Experimental Setup 4.2 Real Datasets and Models 4.3 Analysis of Experimental Results on Synthetic Datasets 4.4 Analysis of Experimental Results on Real Datasets 5 Conclusion References Electricity Energy Monitoring System for Home Appliances Using Raspberry Pi and Node-Red 1 Introduction 2 Methodology 2.1 Planning, Analysis and Design 2.2 System Design 2.3 Block Diagram 2.4 MQTT Connection 3 Result and Discussion 3.1 Variant Comparison 3.2 Programming Result 3.3 Selected Offset Value 3.4 Result Analysis 3.5 Saving Analysis (Tenaga Nasional Berhad) 3.6 Prototype Design 4 Conclusion References Short-Time Fourier Transform with Optimum Window Type and Length: An Application for Sag, Swell and Transient 1 Introduction 2 Power Quality Signal Modelling 3 Windowed Function for STFT 4 Performance Measurements of Windowed STFT 5 Optimization of Window for Reliability, Complexity and Memory 6 Conclusion References Data Analytics and Technologies Cox Point Process with Ridge Regularization: A Better Approach for Statistical Modeling of Earthquake Occurrences 1 Introduction 2 Study Area and Data Description 3 Methodology 3.1 Cauchy Cluster Process 3.2 Intensity Estimation 3.3 Ridge Regularization 3.4 Cluster Parameter Estimation 3.5 Model Interpretation and Selection 4 Result 4.1 Exploratory Data Analysis 4.2 Inference and Model Selection 5 Concluding Remark References Discovering Popular Topics of Sarawak Gazette (SaGa) from Twitter Using Deep Learning 1 Introduction 2 Related Works 2.1 Data Acquisition 2.2 Feature Extraction 2.3 Different Approaches in Topic Modeling 2.4 Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) 2.5 Evaluation 3 Methodology 3.1 Data Acquisition and Data Pre-processing 3.2 Topic Modeling with Latent Dirichlet Allocation 3.3 Classification 3.4 Evaluation 4 Results 5 Discussion 6 Conclusion References Comparison Analysis of LSTM and CNN Variants with Embedding Word Methods for Sentiment Analysis on Food Consumption Behavior 1 Introduction 2 Related Work 3 Material and Methods 3.1 Data Collection 3.2 Case Folding 3.3 Feature Extraction 3.4 Data Labelling 3.5 Model Development 4 Classification Performance Results and Analysis 4.1 Parameter Setting 4.2 Computational Result for CNN 4.3 Computational Result for LSTM 4.4 Computational Result for LSTM-CNN 4.5 Summary of Results 5 Discussion 6 Conclusion References Data Mining and Image Processing Towards Robust Underwater Image Enhancement 1 Introduction 2 Underwater Image Enhancement 2.1 Non-Physical-Based Methods 2.2 Physical-Based Methods 2.3 Deep Learning-Based Methods 3 Methodology 3.1 Dataset 3.2 Evaluation Metrics 4 Result 4.1 Enhancement Result Analysis 4.2 GL-Net Component Analysis 5 Conclusion References A Comparative Study of Machine Learning Classification Models on Customer Behavior Data 1 Introduction 2 Research Methodology 2.1 Dataset Description 2.2 Data Preprocessing 2.3 Model Development 2.4 Model Evaluation 3 Results and Discussion 4 Conclusion References Performance Evaluation of Deep Learning Algorithms for Young and Mature Oil Palm Tree Detection 1 Introduction 2 Background Studies 2.1 Domain Background 2.2 Image Annotation Techniques 2.3 Object Detection Techniques 2.4 Data Splitting Techniques 3 Research Methodology 3.1 Data Collection and Extraction 3.2 Data Pre-processing 3.3 Model Training 3.4 Performance Evaluation 4 Results and Analysis 4.1 Training Data and Model Development 4.2 Model Performance Evaluation 4.3 Parameter Tuning on Best Performed Model 5 Conclusion References Object Detection Based Automated Optical Inspection of Printed Circuit Board Assembly Using Deep Learning 1 Introduction 2 Background Studies 2.1 Object Detection 2.2 Application of Object Detection Using Deep Learning 3 Research Methodology 3.1 Data Collection 3.2 Data Annotation 3.3 Modelling and Evaluation 4 Results and Discussion 4.1 Comparison of Input Image Size 4.2 Training Time 4.3 Accuracy and Resource Trade Off 5 Conclusion References Mathematical and Statistical Learning The Characterization of Rainfall Data Set Using Persistence Diagram and Its Relation to Extreme Events: Case Study of Three Locations in Kemaman, Terengganu 1 Introduction 2 Station and Dataset 3 Kemaman Flood History 4 Topological Data Analysis 4.1 Simplicial Complex 4.2 Homology 4.3 Persistent Homology 5 Sliding window persistence for maximum persistence H1. 5.1 Preprocessing 5.2 Embedding Rainfall Dataset into Higher Dimension 5.3 Persistence Diagram 5.4 Maximum Score 5.5 Confusion Matrix 5.6 Area Under the Receiver Operating Characteristics 6 Results and Discussions 6.1 Stability of Maximum Scores with Missing Value 6.2 Maximum Scores of Rainfalls 7 Conclusion References Clustering Stock Prices of Industrial and Consumer Sector Companies in Indonesia Using Fuzzy C-Means and Fuzzy C-Medoids Involving ACF and PACF 1 Introduction 2 Literature Review 2.1 Fuzzy Clustering 2.2 Fuzzy Silhouette 3 Methodology 3.1 Data Set 3.2 Methods 4 Results and Discussions 4.1 Simulation Study 4.2 Clustering Open Stock Prices 4.3 Clustering Close Stock Prices 4.4 Clustering HML Stock Prices 4.5 Comparison of Clustering Results 5 Conclusions References Zero-Inflated Time Series Model for Covid-19 Deaths in Kelantan Malaysia 1 Introduction 2 Related Works 3 Material and Methods 3.1 Material 3.2 Predictive Models for Count Data 3.3 Model Evaluation 3.4 Vuong Test 3.5 Incidence Rate Ratios 4 Result and Discussion 4.1 Descriptive Analysis 4.2 Analysis Basic Count Models 4.3 Variables Selection 4.4 Model Comparison between Reduced Zero-Inflated Models 5 Conclusion and Future Work References Author Index