Machine Learning and Knowledge Extraction: 5th IFIP TC 5, TC 12, WG 8.4, WG 8.9, WG 12.9 International Cross-Domain Conference, CD-MAKE 2021, Virtual ... (Lecture Notes in Computer Science, 12844)

This book constitutes the refereed proceedings of the 5th IFIP TC 5, TC 12, WG 8.4, WG 8.9, WG 12.9 International Cross-Domain Conference, CD-MAKE 2021, held in virtually in August 2021. The 20 full papers and 2 short papers presented were carefully reviewed and selected from 48 submissions. The cross-domain integration and appraisal of different fields provides an atmosphere to foster different perspectives and opinions; it will offer a platform for novel ideas and a fresh look on the methodologies to put these ideas into business for the benefit of humanity. Preface Organization Contents Digital Transformation for Sustainable Development Goals (SDGs) - A Security, Safety and Privacy Perspective on AI 1 Introduction and Motivation 2 Background and Related Work 2.1 The UN Sustainable Development Goals (SDGs) 2.2 Smart Farming and Precision Agriculture - An Example for Supporting SDGs Through AI 2.3 Smart Health and Precision Medicine - An Example for Supporting SDGs Through AI 2.4 Impact of AI on SDGs 3 Open Issues on Using AI for SDGs 3.1 Data and Models 3.2 Providing Trustworthy Systems 3.3 Control 3.4 Transparency 3.5 Other Issues 4 Conclusion References When in Doubt, Ask: Generating Answerable and Unanswerable Questions, Unsupervised 1 Introduction 1.1 Problem Statement 1.2 Prior Research 1.3 Objective and Contributions 2 Method 2.1 Proposed Model 2.2 Dataset 2.3 Evaluation Metrics 3 Experiments 4 Results and Discussion 5 Conclusions 6 Limitations 7 Future Work References Self-propagating Malware Containment via Reinforcement Learning 1 Introduction 2 Related Work 3 Methods 3.1 Environment 3.2 Feature Selection 3.3 Agent 4 Experiments 5 Discussion 6 Conclusion References Text2PyCode: Machine Translation of Natural Language Intent to Python Source Code 1 Introduction 2 Proposed Approach 3 Dataset 3.1 Extraction and Preparation 3.2 Dataset Description 4 Experiments 4.1 Baseline Model 4.2 Metrics 5 Results 6 Conclusion References Automated Short Answer Grading Using Deep Learning: A Survey 1 Introduction 2 Methodology 3 Corpora 4 Evaluation Metrics 4.1 Regression Metrics 4.2 Classification Metrics 4.3 F1-Score 5 Deep Learning Approaches 6 Results 7 Conclusion and Future Challenges References Fair and Adequate Explanations 1 Introduction 2 Background on Explanations 2.1 Counterfactual Explanations for Learning Algorithms 3 From Partial to More Complete Explanations 4 Pragmatic Constraints on Explanations 5 The Algorithmic Complexity of Finding Fair and Adequate Explanations 6 Conclusion References Mining Causal Hypotheses in Categorical Time Series by Iterating on Binary Correlations 1 Introduction 2 Methodology 2.1 Program 2.2 Data Sources 2.3 Data Model 2.4 Correlation Analysis 2.5 Causality Analysis 2.6 Result Representation 3 Conclusion References Active Finite Reward Automaton Inference and Reinforcement Learning Using Queries and Counterexamples 1 Introduction 1.1 Related Works 2 Finite Reward Automata 2.1 Markov Decision Processes and Finite Reward Automata 2.2 Reinforcement Learning with Finite Reward Automata 3 Expressivity of Finite Reward Automata 4 Active Finite Reward Automaton Inference and Reinforcement Learning (AFRAI-RL) 4.1 Active Finite Reward Automaton Inference Engine 4.2 Active Reinforcement Learning Engine 5 Case Studies 5.1 Office World Scenario 5.2 Minecraft world scenario 6 Conclusions References Rice Seed Image-to-Image Translation Using Generative Adversarial Networks to Improve Weedy Rice Image Classification 1 Introduction 2 Related Work 2.1 Rice Classifications 2.2 Generative Adversarial Networks 2.3 Conditional Adversarial Networks 2.4 Image-to-Image Translation 2.5 Unpaired Image-to-Image Translation 3 Dataset 4 Approach 4.1 Main Components 4.2 Objective Formulation 4.3 Model Architecture 4.4 Training Details 5 Evaluations 5.1 Benchmarks 5.2 Validation Set 5.3 Test Set 6 Conclusions References Reliable AI Through SVDD and Rule Extraction 1 Introduction 2 Support Vector Data Description 2.1 Theory 2.2 Autonomous Detection of SVDD Parameters with RBF Kernel 2.3 Fast Training SVDD 2.4 Zero FNR Regions with SVDD 3 Rules Extraction 3.1 Logic Learning Machine 3.2 Rules Extraction from SVDD 4 Applications: The Vehicle Platooning Example 4.1 Vehicle Platooning 5 Conclusions and Future Works References Airbnb Price Prediction Using Machine Learning and Sentiment Analysis 1 Introduction 2 Related Work 3 Dataset 3.1 Sentiment Analysis on the Reviews 3.2 Feature Selection 4 Methods 4.1 Ridge Regression 4.2 K-means Clustering with Ridge Regression 4.3 Support Vector Regression 4.4 Neural Network 4.5 Gradient Boosting Tree Ensemble 5 Experiments and Discussion 6 Conclusions and Future Work A Appendix References Towards Financial Sentiment Analysis in a South African Landscape 1 Introduction 2 Sentiment Analysis and Opinion Mining 3 Financial Sentiment Analysis 3.1 Exploiting Typical Financial Headline Structure 3.2 Existing Approaches for Financial Sentiment Analysis 3.3 Related Work on Financial Sentiment Analysis in the South African Context 4 Sentiment Correlation with Financial Performance 5 Method 5.1 Topic Modelling for Data Filtering 5.2 Annotation of Data 5.3 Sentiment Prediction Model 5.4 Correlation with Financial Performance 6 Data 7 Model Development Results 7.1 Annotation of Data 7.2 Sentiment Prediction Model 7.3 Sentiment Correlation with Financial Performance 8 Model Generalisation 9 Conclusions and Future Work References Weighted Utility: A Utility Metric Based on the Case-Wise Raters' Perceptions 1 Introduction 2 Methods 2.1 Weighted Utility Metrics 2.2 Experimental Evaluation 3 Results and Discussion 4 Conclusions References Deep Convolutional Neural Network (CNN) Design for Pathology Detection of COVID-19 in Chest X-Ray Images 1 Introduction 1.1 Initial Goals 2 Data Sources 3 Data Integration and Preprocessing 3.1 Applying Transfer Learning for Classifying X-Ray Images 3.2 Improving Model Performance Using Darknet-53(Yolo-V3) 4 Results 4.1 Model Performance Evaluation Using Statistical Methods 5 Discussion and Conclusion 5.1 Future Directions 5.2 Limitations References Anomaly Detection for Skin Lesion Images Using Replicator Neural Networks 1 Introduction 2 Related Work 3 Method 4 Results 5 Conclusions References On the Overlap Between Grad-CAM Saliency Maps and Explainable Visual Features in Skin Cancer Images 1 Introduction 2 Related Work 3 Classification Architectures and Models 4 Data Preparation 5 First Experiment 6 Second Experiment 7 Discussion 8 Conclusions and Future Work References From Explainable to Reliable Artificial Intelligence 1 Introduction 2 Related Work 2.1 Safety Engineering-Based Methods 2.2 Classification with Abstension 2.3 Explainable AI-Based Methods 3 Logic Learning Machine 3.1 Feature and Value Ranking 4 Skope-Rules 5 Reliability Assessment Methods 5.1 Reliability from Outside 5.2 Reliability from Inside 5.3 Rules with Zero Error 6 Applications and Results 6.1 Physical Fatigue 6.2 Vehicle Platooning 6.3 Discussion 7 Conclusions and Future Works References Explanatory Pluralism in Explainable AI 1 Introduction 2 Explanatory Demands 2.1 Fulfilling Disparate Explanatory Demands 3 Scientific Explanations 3.1 Manipulationist Account of Causation 3.2 Mechanistic Account of Scientific Explanations 4 Pluralistic Taxonomy 4.1 Three Levels of Analysis (Plus One) 4.2 Mechanistic-Social, Particular-General Taxonomy 5 Pragmatic Interventionist Stance 5.1 Organizing Present XAI Methods 5.2 Descriptive vs Evaluative Taxonomy 6 Conclusion 6.1 Limit of Diagnostics 6.2 Ratiocinative AI 6.3 On Firmer Grounds References On the Trustworthiness of Tree Ensemble Explainability Methods 1 Introduction 2 Background 3 Experimental Setup 4 Results 4.1 Accuracy of Gain and SHAP Feature Importances 4.2 Stability of Gain and SHAP Feature Importances 4.3 Summary of Results 5 Discussion A Determinism of XGBoost Feature Importances References Human-in-the-Loop Model Explanation via Verbatim Boundary Identification in Generated Neighborhoods 1 Introduction 2 Related Work 3 The Proposed Human-in-the-Loop Framework 3.1 Stage (I): Neighborhood Generation 3.2 Stage (II): Neighborhood Classification 3.3 Stage (III): Human-in-the-Loop Exploration 4 Experiment Setup 4.1 Dataset and Trained Machine Learning Architecture 4.2 Our Proposed Framework Settings 5 Result 5.1 MNIST 5.2 FashionMNIST 6 Workflow of Human Users of the Proposed Framework 6.1 Identifying the Point-of-Interest 6.2 Identifying Interesting Dimensions and Appropriate Step Lengths 6.3 Selecting Two Most Revealing Dimensions to Generate a Matrix for Decision Boundary Visualization 7 Discussion, Limitations and Future Works 8 Conclusion References MAIRE - A Model-Agnostic Interpretable Rule Extraction Procedure for Explaining Classifiers 1 Introduction 2 Related Work 3 Methodology 3.1 Problem Statement 3.2 Approximations to Coverage and Precision 3.3 Accuracy of the Approximation 4 Optimization to Estimate the Explanation 4.1 Greedy Attribute Elimination for Human Interpretability 4.2 Local to Global Explanations 4.3 Extension to Discrete Attributes 5 Experiments and Results 5.1 Tabular Datasets 5.2 Text Datasets 5.3 Image Datasets 6 Summary References Transparent Ensembles for Covid-19 Prognosis 1 Introduction 2 Models 2.1 Ensembles of Binary Decision Trees 2.2 Transformation of Trees into Transparent MLPs 2.3 DIMLP Networks 3 Experiments 3.1 Models and Learning Parameters 3.2 Random Forest Results 3.3 Gradient Boosting Results 3.4 Results Obtained by DIMLPs Trained by Bagging 3.5 Results of Boosted Decision Stumps 3.6 An Example of Generated Ruleset 3.7 Related Work 4 Conclusion References Author Index