Contents
Data Science and Knowledge Discovery Using Machine Learning Methods
1 Introduction
2 The KDD Process
3 Taxonomy of Data Science Methods
4 Data Science Within the Complete Decision Support System
5 KDD and Data Science Research Opportunities and Challenges
6 KDD and DM Trends
7 The Organization of the Handbook
7.1 Data Preparation Methods
7.2 Supervised Learning
7.3 Unsupervised Learning Methods
7.4 Deep Learning
7.5 Methods for Special Data Setting
7.6 Methods for Special Learning Tasks
7.7 Domain-Specific Applications
7.8 Human Factors and Social Issues
Handling Missing Attribute Values
1 Introduction
2 Sequential Methods
2.1 Deleting Cases with Missing Attribute Values
2.2 The Most Common Value of an Attribute
2.3 The Most Common Value of an Attribute Restricted to a Concept
2.4 Assigning All Possible Attribute Values to a Missing Attribute Value
2.5 Assigning All Possible Attribute Values Restricted to a Concept
2.6 Replacing Missing Attribute Values by the Attribute Mean
2.7 Replacing Missing Attribute Values by the Attribute Mean Restricted to a Concept
2.8 Global Closest Fit
2.9 Concept Closest Fit
2.10 Other Methods
3 Parallel Methods
3.1 Blocks of Attribute-Value Pairs and Characteristic Sets
3.2 Lower and Upper Approximations
3.3 Rule Induction—MLEM2
3.4 Other Approaches to Missing Attribute Values
4 Conclusions
References
Data Integration Process Automation Using Machine Learning: Issues and Solution
1 Introduction
2 Related Work
3 Case Study
3.1 Manufacturing Industry
3.2 Insurance Industry
3.3 Banking Industry
3.4 Aviation Industry
4 Proposed Solution
4.1 Automated Data Integration
4.2 Details of Major Components
4.2.1 Database Version Control
4.2.2 Custom Rule Engine
4.2.3 Data Pre-processor
4.2.4 Database Release Automation
5 Solutions to Case Study Problems
5.1 Case Study 1: Manufacturing
5.2 Case Study 2: Insurance
5.3 Case Study 3: Banking
5.4 Case Study 4: Aviation
6 Conclusion and Future Study
References
Rule Induction
1 Introduction
2 Types of Rules
3 Rule Induction Algorithms
3.1 LEM1 Algorithm
3.2 LEM2 Algorithm
3.3 MLEM2 Algorithm
3.4 AQ Algorithm
4 Classification Systems
5 Validation
6 Advanced Methodology
References
Nearest-Neighbor Methods: A Modern Perspective
1 Introduction
1.1 Metric Spaces
1.2 Learning Framework
1.2.1 Consistency and Typical Rates
1.3 Computational Efficiency
2 Vanilla Nearest-Neighbor Prediction
2.1 A Link Between Regression and Classification
2.2 Choice of Distance ρ(x, x')
2.2.1 Choices of ρ in Practice
2.2.2 Theoretical Insights on Choices of ρ
2.3 Choice of k = k(n)
3 Distributed 1-NN vs. Weighted k-NN
4 Compression-Based 1-NN
References
Support Vector Machines
1 Introduction
2 Hyperplane Classifiers
2.1 The Linear Classifier
2.2 The Kernel Trick
2.3 The Optimal Margin Support Vector Machine
3 Non-separable SVM Models
3.1 Soft Margin Support Vector Classifiers
3.2 Support Vector Regression
3.3 SVM-Like Models
4 Implementation Issues with SVM
4.1 Optimization Techniques
4.2 Model Selection
4.3 Multi-Class SVM
5 Extensions and Application
6 Conclusion
References
Empowering Interpretable, Explainable Machine Learning Using Bayesian Network Classifiers
1 Introduction
2 Restricted BNCs–The Naïve Bayesian Classifier and Its Variants
2.1 Naive Bayesian Classifier
2.2 Variants of the NBC
2.3 Experimental Evaluation of NBC Variants
3 Beyond the NBC—the Unrestricted BNC
3.1 The General BN
3.2 The General BNC
3.3 Risk Minimization by Cross-Validation
3.4 Experimental Evaluation of Unrestricted BNCs
3.4.1 Evaluation of BNCs Using Traditional Data Sets
3.4.2 Evaluation of BNCs Using Authentic Data Sets
4 Beyond the BNC–Causal–Temporal Classifiers
5 Conclusion and Discussion
References
Soft Decision Trees
1 Introduction
1.1 Research Motivation and Direction
1.2 Organization of the Work
2 Soft Probability: Incorporation of Soft Number into Probability Theory
2.1 Examples on Mixture Models
2.2 Observations
3 Soft Decision Trees
3.1 Overview
3.2 Soft Decision Trees Based on C4.5 Algorithm
3.2.1 Preferring a Thesis in Equilibrium State
3.2.2 Example: Uniformly Distribution Features and Discrete Label
3.2.3 Example: Electric Product Warranty
4 Conclusions
5 Suggestions for Future Research
Appendix 1: Complement, Union, Intersection, and Conditional Soft Probabilities
Complement, Unions, and Intersections
Conditional Probability
Extension of Soft Probability for 2 Dimensions
Appendix 2: Presentation of Soft Numbers
Soft Number: Definitions and Axioms
Mathematical Operations and Functions on Soft Numbers
Soft Axis Coordinate System
References
Quality Assessment and Evaluation Criteria in Supervised Learning
1 Introduction
1.1 Notations and Definitions
2 Hard-Decision Classifiers
2.1 The Confusion Matrix
2.2 The Binary Confusion Matrix
2.3 Performance Measures
2.3.1 Accuracy and Error Rate
2.3.2 True-Positive Rate, False-Positive Rate, and Likelihood Ratios
2.3.3 Positive- and Negative-Predictive Values
2.3.4 Precision, Recall, and F-measure
2.3.5 Choosing a Hard-Decision Measure
2.4 Skew and Cost Considerations
3 Scoring Classifiers
3.1 ROC and AUC
3.2 Additional Graphical Measures
4 Probabilistic Classifiers
4.1 Proper Performance Measures
4.2 Regret and Divergence
4.3 Loss Functions and Performance Measures
4.4 Comparing Probabilistic Performance Measures
4.5 Universal Performance Measures
4.6 A Real-World Example
5 Regression Problems
5.1 Mean Square Error
5.1.1 R-Squared
5.2 Mean Absolute Value
5.3 Percentage Error
6 Discussion
References
Trajectory Clustering Analysis
1 Introduction
2 Preliminaries
2.1 Summary of Main Notations
2.2 Problem Statement
3 Trajectory Clustering via Atomic Representation
3.1 Atomic Representation
3.2 Algorithm of ARSC
3.3 Related Works
4 Minimum Error Entropy-Based Trajectory Clustering Analysis
4.1 Minimum Error Entropy
4.2 MEESSC
5 Experiments
5.1 Experimental Settings
5.2 Motion Segmentation
5.3 Face Clustering
5.3.1 Face Clustering Using Extended Yale B Database
5.3.2 Face Clustering with Contiguous Occlusion
5.3.3 Face Clustering with Missing Entries
6 Conclusions
References
Clustering High-Dimensional Data
1 Introduction
2 Axis-Parallel Subspace Clustering
2.1 Subspace Search Strategy
2.2 Cluster Criteria
3 Arbitrarily Oriented Subspace Clustering
3.1 Principal Component Analysis (PCA)
3.2 Grid-Based Cluster Identification
3.3 Hough Transform
3.4 Random Sampling
3.5 Intrinsic Dimensionality
4 Conclusion
4.1 Further Reading
4.2 Future Research Directions
References
Fuzzy C-Means Clustering: Advances and Challenges (Part II)
1 Introduction
2 Structure of Classical FCM
2.1 Variants and Advancements of FCM
2.1.1 Intuitionistic FCM (IFCM)
2.1.2 Fuzzy Possibilistic C-Means
2.1.3 Kernel-Based FCM
2.1.4 Type-2 FCM (T2FCM)
2.1.5 Interval Type-2 FCM (IT2-FCM)
2.1.6 Hierarchical FCM (HFCM)
2.1.7 Robust FCM
2.1.8 Semi-supervised FCM
2.1.9 Hybrid FCM
3 Applications of FCM and Its Variants
3.1 FCM and Neural Networks
3.2 Image Analysis Using FCM
3.3 Clustering and Classification Using FCM
3.4 Intrusion Detection Using FCM
4 Critical Analysis
5 Conclusions
References
Clustering in Streams
1 Introduction
2 Representative-Based Methods
2.1 The STREAM Algorithm
2.2 Micro-clustering: The Big Picture
2.2.1 Defining Micro-clusters
2.2.2 Pyramidal Time Frame
2.2.3 How Micro-clusters Are Created in Real Time
3 Density-Based Stream Clustering
3.1 DenStream: Density-Based Micro-clustering
3.2 Grid-Based Streaming Algorithms
3.2.1 D-Stream Algorithm
3.2.2 Other Grid-Based Algorithms
4 Probabilistic Streaming Algorithms
5 Clustering High-Dimensional Streams
5.1 The HPSTREAM Method
5.2 Other High-Dimensional Streaming Algorithms
6 Clustering Categorical Streams
6.1 Clustering Binary Data Streams with k-Means
6.2 The StreamCluCD Algorithm
6.3 Massive-Domain Clustering
7 Text Stream Clustering
8 Other Scenarios for Stream Clustering
8.1 Clustering Uncertain Data Streams
8.2 Clustering Graph Streams
8.3 Distributed Clustering of Data Streams
9 Discussion and Conclusions
References
Introduction to Deep Learning
1 General Overview
2 Basic NN Structure
2.1 Common Linear Layers
2.2 Common Nonlinear Functions
3 Loss Functions
3.1 Metric Learning
4 Neural Network Training
4.1 Backpropagation
4.2 Training Considerations
4.3 Hyper-Parameter Tuning
5 Training Optimizers
6 Training Regularizations
7 Advanced NN Architectures
7.1 Deep Learning for Detection and Segmentation
7.2 Deep Learning on Sequential Data
7.3 Deep Learning on Irregular Grids
8 Summary
References
Graph Embedding
1 Introduction
2 Definitions
3 Static Graph Embedding Methods
3.1 Graph Properties Preserved
3.1.1 Community Preserving Graph Embedding Methods
3.1.2 Structure Preserving Graph Embedding Methods
3.2 Manifold Learning
3.3 Model Scalability
4 Dynamic Graph Embedding Methods
4.1 Updating Graph Embedding
4.2 Capturing Temporal Patterns
5 Attributed Graph Embedding Methods
6 Conclusion
References
Autoencoders
1 Autoencoders
2 Regularized Autoencoders
2.1 Sparse Autoencoders
2.2 Denoising Autoencoders
2.3 Contractive Autoencoders
3 Variational Autoencoders
3.1 The Reparameterization Trick
3.2 Example: The Case of Normal Distribution
3.3 Disentangled Autoencoders
4 Applications of Autoencoders
4.1 Autoencoders as a Generative Model
4.2 Use of Autoencoders for Classification
4.3 Use of Autoencoders for Clustering
4.4 Use of Autoencoders for Anomaly Detection
4.5 Use of Autoencoders for Recommendation Systems
4.6 Use of Autoencoders for Dimensionality Reduction
5 Advanced Autoencoder Techniques
5.1 Autoencoders and Generative Adversarial Networks
5.2 Adversarially Learned Inference
5.3 Wasserstein Autoencoders
5.4 Deep Feature Consistent Variational Autoencoder
5.5 Conditional Image Generation with PixelCNN Decoders
6 Conclusion
References
Generative Adversarial Networks
1 Introduction to GANs
2 The Basic GAN Concept
3 GAN Advantages and Problems
3.1 Mode Collapse
3.2 Vanishing Gradients
3.3 Instability and Image Quality
3.4 Problems: Summary
4 Improved GAN Architectures
4.1 Semi-Supervised GAN (SGAN)
4.2 Conditional GAN (CGAN)
4.3 Deep Convolutional GAN (DCGAN)
4.4 Progressive GAN (PROGAN)
4.5 BigGAN
4.6 StyleGAN
5 Improved GAN Objectives
5.1 Wasserstein GAN (WGAN)
5.2 Self-Supervised GAN (SSGAN)
5.3 Spectral Normalization GAN (SNGAN)
5.4 SphereGAN
6 Data Augmentation with GAN
7 Conclusion
References
Spatial Data Science
1 Introduction
2 Spatial Data
2.1 Types of Spatial Data
2.2 Major Sources of Spatial Data
3 Spatial Database
4 Spatial Data Mining
4.1 Spatial Hotspots
4.2 Co-locations
4.3 Spatial Outliers
4.4 Spatial Prediction
4.5 Spatial Change
5 Result Validation
5.1 Key Concepts
5.2 Formulations and Computational Techniques
5.2.1 Spatial Hotspots
5.2.2 Other Patterns
6 Research Trend
7 Conclusion
References
Multimedia Data Learning
1 Introduction
2 A Typical Architecture of a Multimedia Data Mining System
3 Feature Extraction
4 Knowledge Representation
5 Learning Methodology
5.1 Statistical Learning
5.2 Soft Computing Based Learning
6 Conclusion and Further Readings
References
Web Mining
1 Introduction
2 The Graph Structure of the Web
3 Web Content Mining
4 Web Structure Mining
5 Web Usage Mining
6 The Semantic Web and Semantic Web Mining
References
Mining Temporal Data
1 Introduction
2 Time Point Series Mining
3 Classification with Time Series
4 From Time Point Series to Symbolic Temporal Data
5 Mining Sequential Data
6 Mining Time Intervals Data
7 Temporal Patterns Based Classification
8 Discussion
References
Cloud Big Data Mining and Analytics: Bringing Greenness and Acceleration in the Cloud
1 Introduction
2 Big Data Mining and Analytics in the Cloud
3 Graphical Processing Units (GPUs) and Cloud Big Data Analytics
4 Approximate Computing (AC) and Cloud Big Data Analytics
5 Quantum Computing (QC) and Cloud Big Data Analytics
6 Discussions, Prospects, and Future Trends
7 Conclusions
References
Multi-Label Ranking: Mining Multi-Label and Label Ranking Data
1 Introduction
2 Definition and Context
3 Multi-Label Algorithms
3.1 Problem Adaptation and Problem Transformation
3.2 Multi-Label Ensembles
3.3 Deep Learning Methods
3.3.1 Image Annotation
3.3.2 Text Annotation
3.4 Extreme Multi-Label Classification
4 Label Ranking Algorithms
4.1 Label Ranking Ensembles
5 Evaluation
5.1 Dataset Repositories
5.2 Stratification of Multi-Label Data
6 Research Directions and Open Problems
References
Reinforcement Learning for Data Science
1 Introduction
2 Reinforcement Learning Formulation
2.1 Problem Formulation
2.2 Bellman Equation
2.3 Q-Learning and SARSA
2.4 Exploration-Exploitation
2.5 On- and Off-Policy
2.6 Actor-Critic
3 Curious Feature Selection
3.1 Feature Selection
3.2 Intrinsically Motivated Learning and the Curiosity Loop
3.3 States and Actions
3.4 Learner and Internal Reward
3.5 State-Action Value Function and Policy
3.6 Exploration vs. Exploitation
3.7 Experimental Setting
3.7.1 Diabetes Dataset
3.7.2 Data Preparation
3.7.3 Comparison Algorithms
3.7.4 Results
3.8 Summary
4 Deep Reinforcement Learning
4.1 Introduction to DeepRL
4.2 Difficulties Relevant to DeepRL
4.2.1 Sample Efficiency
4.2.2 Sparse Rewards
4.2.3 Correlation between Data Points
4.2.4 Ground Truth Unavailability
4.2.5 Benchmarks
4.3 DeepRL Algorithms
4.3.1 Deep Q-Networks (DQN)
4.3.2 Q-Networks and Target Networks
4.3.3 Experience Replay
5 Summary
References
Adversarial Machine Learning
1 Introduction
2 The Very Early Works
3 The Evolution of White-Box Attack & Defense Methods against Deep Neural Networks
3.1 Preliminaries and Notations
3.1.1 Notations
3.1.2 Targeted and Non-targeted Attack Methods
3.1.3 Measuring the Magnitude of Perturbations
3.2 The Basic Recipe for a White-Box Attack against a Neural Network
3.3 Early Attack Methods
3.3.1 Fast Gradient Sign Method—FGSM
3.3.2 Jacobian Saliency Map Attack—JSMA
3.3.3 Targeted Gradient Sign Method—TGSM
3.3.4 Basic Iterative Method—BIM
3.4 Early Defense Mechanisms
3.4.1 Adversarial Retraining
3.4.2 Defensive Distillation
3.5 The Carlini & Wagner Attack—Heralding a Second Generation of Attack Methods
3.6 Additional (Unsuccessful) Defense and Detection Methods
3.6.1 GAN-Based Defenses
3.6.2 Detecting Adversarial Examples
4 Black-Box Attack Methods
4.1 Adversarial Transferability
4.2 Zeroth Order Optimization-Based Attack
4.3 The HopSkipJump Attack
5 Real World Adversarial Attacks
6 Theoretical Reasoning and Outstanding Research Questions
6.1 Adversarial Examples as a Model Generalization Problem
6.2 Adversarial Examples are Inevitable
6.2.1 Inevitability of Existence
6.2.2 Inevitability of Finding
6.3 Additional Open Research Subjects
7 Summary
References
Ensembled Transferred Embeddings
1 Introduction
2 The ETE Framework
3 Applying ETE to the Item Categorization Task
3.1 Datasets
3.1.1 Invoice Dataset
3.1.2 eBay Dataset
3.2 Sample Dataset—Labeling Using MTurk
3.3 Transferred Embeddings—Using Seller's Industry Attribute and eBay Dataset
3.4 Transferred Models—Using a Fully Connected Neural Network
3.5 Ensemble—Stacking Models
4 Evaluation
4.1 Experimental Setting
4.1.1 Compared Methods
4.1.2 Dataset and Pre-processing
4.1.3 Hyper-parameter Tuning
4.2 Results
5 Summary and Future Work
References
Data Mining in Medicine
1 Introduction
2 Machine Learning and Some Emerging Medical Applications
2.1 Clinical Imaging
2.1.1 Diabetic Retinopathy (DR)
2.1.2 Breast Cancer and Lymph node Metastasis
2.1.3 Skin Cancer
2.2 Electronic Medical Records
2.3 Challenges in Healthcare
3 Temporal Data Mining in Medicine: Some Relevant Techniques
3.1 Temporal Abstractions and the Knowledge-Based Temporal Abstraction (KBTA) Method
3.2 Association Rules
3.3 Temporal Pattern Mining
3.4 Approximate Temporal Functional Dependencies
3.4.1 Temporal Functional Dependencies (TFD)
3.4.2 Approximate Functional Dependencies (AFD)
3.4.3 Approximate Temporal Functional Dependencies (ATFD)
4 Conclusions
References
Recommender Systems
1 Introduction to Recommender Systems
1.1 Concepts and Notations
2 Recommendation Techniques
2.1 Non-personalized and Lightly Personalized Recommendations
2.2 Neighborhood Methods
2.2.1 User-Based Collaborative Filtering
2.2.2 Item-Based Collaborative Filtering
2.3 Factorization-Based Methods
2.3.1 Matrix Factorization
2.3.2 Factorization Machines
2.3.3 Collaborative Metric Learning
2.4 Modeling Sequences in Recommendation
2.5 Neural Architectures for Recommender Systems
2.5.1 From Linear to Nonlinear Recommendation Models
2.5.2 Representation Learning with Neural Architectures
2.5.3 Sequence-Aware Recommendation with Neural Networks
2.5.4 Advanced Topics and New Frontiers
2.6 Content-Based Recommender Systems
3 Recommendation Tasks and Applications
3.1 Applications of Recommender Systems
3.2 Practice for Industrial-Scale Recommendation
3.3 Tool Kits for Building Recommender Systems
4 Evaluate Recommender Systems
4.1 Evaluation on Recommendation Accuracy
4.2 Beyond Accuracy
5 Conclusion
References
Activity Recognition
1 Introduction
2 The Procedure of Activity Recognition
2.1 Problem Formulation
2.2 Sensor Inputs
2.3 Feature Engineering
2.4 Model
2.5 Evaluation
3 Hot Research Areas
3.1 Transfer Learning Based HAR
3.2 Federated Learning Based HAR
3.3 Incremental Learning Based HAR
4 Datasets
5 Challenges
5.1 Online and Mobile Deep Activity Recognition
5.2 More Accurate Unsupervised Activity Recognition
5.3 Flexible Models to Recognize High-Level Activities
5.4 Light-Weight Deep Models
5.5 Non-invasive Activity Sensing
5.6 Beyond Activity Recognition: Assessment and Assistant
6 Conclusions
References
Social Network Analysis for Disinformation Detection
1 Introduction
2 Related Work
2.1 News Detection Based on Diffusion Analysis
2.2 Fake News Detection Based on Sources
2.3 Detecting Fake News Based on Online Social Media Features
3 Methods
3.1 Data Collection
3.1.1 Claim Collection
3.1.2 Keyword Assignment
3.1.3 Post and Author Collection
3.2 Feature Extraction
3.2.1 Author-Based Features
3.2.2 Post-Based Features
3.2.3 Fake News Oriented Word-Based Features
3.2.4 Word Embedding-Based Features
3.2.5 Doc2vec-Based Features
3.2.6 Topic-Based Features
3.2.7 Sentiment-Based Features
3.2.8 Temporal Features
3.2.9 Behavioral Features
3.2.10 Graph-Based Features
3.3 Classification
4 Data
4.1 Claims
4.2 Keywords
4.3 Posts and Authors
5 Evaluation
5.1 Experimental Setup
5.2 Results and Discussion
5.2.1 Feature Importance
5.2.2 Best Classifier
5.3 Feature Subset Comparison
5.3.1 Number of Posts per Claim
6 Ethical Considerations
7 Conclusion and Future Work
References
Online Propaganda Detection
1 Introduction
2 Terrorist Content Detection
2.1 Online Terrorist Propaganda
2.2 Using Word Graphs for Terrorist Content Detection
2.3 Case Studies
2.4 Using Hybridized Term-Weighting Method for Terrorist Content Detection
2.5 Using Linguistic Markers for Detecting Violent Online Content
3 Detecting Propaganda Campaigns on Social Media Platforms
3.1 Using Targeted Advertising on Facebook as a Propaganda Tool
3.2 Disinformation Campaigns on Twitter
3.3 Fake News Detection
3.4 Troll Accounts Detection
3.5 Propaganda and Misinformation in Social Media
4 Conclusion and Future Research
References
Interpretable Machine Learning for Financial Applications
1 Introduction: Financial Tasks
2 Methodologies
2.1 Specifics of Machine Learning in Finance
2.2 Aspects of ML Methodology in Finance
3 ML Models and Practice in Finance
3.1 Portfolio Management and Neural Networks
3.2 Interpretable Trading Rules and Relational ML
3.3 Relational Machine Learning for Trading
3.4 Visual Knowledge Discovery for Currency Exchange Trading
3.5 Discovering Money Laundering and Attribute-Based Relational Machine Learning
4 Conclusion
References
Predictive Analytics for Targeting Decisions
1 Introduction
2 PA – Predictive Analytics
3 How to Evaluate a PA Model?
4 How to Select the Explanatory Variables for the Model?
5 Explanatory Model Versus Prediction Model
6 So How Does One Build a Good Prediction Model?
7 Performance Measures for Assessing Prediction Accuracy
8 How to Detect Over-Fitting?
9 Economic Considerations and Decision-Making
10 Using Decision Trees for Feature Selection and Prediction
11 Conclusions
References
Machine Learning for the Geosciences
1 Introduction
2 Background on Seismology
3 A General Machine Learning Framework for Seismic Signal Processing
4 Seismic Event Detection and Localization
5 Seismic Event Classification
6 Deep Learning for Seismic Signal Processing
6.1 Seismic Event Detection Using Deep Learning Techniques
6.2 Seismic Event Localization and Characterization with Deep Learning Models
7 Conclusions
References
Sentiment Analysis for Social Text
1 Introduction
1.1 Ontology
2 Text Representation and Normalization
2.1 Resources
3 Lexical-Level Sentiment Analysis
3.1 Resources
4 Aspect-Level Sentiment Analysis
4.1 Unsupervised
4.2 Weakly Supervised
4.3 Supervised
4.4 Resources
5 Sentence-Level Sentiment Analysis
5.1 Resources
6 Applications
7 Summary and Discussion
7.1 A Call for Responsible SA
References
Human Resources-Based Organizational Data Mining (HRODM): Themes, Trends, Focus, Future
1 Introduction
2 Methodology
2.1 Database Development
2.2 Categorization
2.3 Classification System
3 Results and Discussion
3.1 Emergence of HRODM Research
3.2 Trends in HRODM Research
3.3 Theoretical Framework: ROI-Based Analysis of HRODM
3.4 Empirical Studies
3.5 Conceptual Studies
3.6 Case-Based Studies
3.7 Technical Studies
4 HRODM: Practical Implementation Tools and Expected ROI
4.1 Implications for Organizations
5 Contributions: ROI – Model to Guide the Way Forward
6 HRODM Implementation: Two Examples
6.1 HRODM Implementation Example I: Workforce Analytics and Big Data Analysis of Employee Turnover
6.2 HRODM Implementation Example II: Employee Recruitment Decision-Making Support Tool
References
Algorithmic Fairness
1 Introduction
2 Potential Causes of Unfairness
3 Fairness Definitions and Measures
3.1 Definitions of Discrimination in Legal Domains
3.2 Measures of Algorithmic Bias
3.3 Trade-Offs
4 Fairness-Enhancing Mechanisms
4.1 Pre-process Mechanisms
4.2 In-process Mechanisms
4.3 Post-process Mechanisms
4.4 Which Mechanism to Use?
5 Fairness-Related Datasets
6 Discussion and Conclusion
References
Privacy-Preserving Data Mining (PPDM)
1 Introduction
2 PPDM Classification
2.1 Data Attributes Classification
3 Anonymization
3.1 k-Anonymity
3.2 l-Diversity
3.3 t-Closeness
4 Randomization
4.1 Perturbation
4.2 Multiplicative Noise
4.3 Differential Privacy (DP)
4.4 Data Swapping
5 Cryptography
5.1 Secure Multi-party Computation
5.2 Homomorphic Encryption
6 Privatizing Data Mining Results
6.1 Association Rule Hiding
6.2 Downgrading Classifiers Accuracy
6.3 Query Auditing and Inference Control
7 Utility Assessment
8 Personalized Privacy and Risk Assessment
9 Conclusions
References
Explainable Machine Learning and Visual Knowledge Discovery
1 Introduction
1.1 Motivation
1.2 Analytical Versus Visual ML
1.3 Approaches
2 Visualization of Analytical ML Models
2.1 Matrix and Parallel Sets Visualization for Association Rules
2.2 Dataflow Visualization in ML Models
2.3 Input-Based ML Model Visualization for Binary Data
3 Discovery of Analytical ML Models Aided by Visual Methods
3.1 Approaches to Visualize Multidimensional Data
3.2 Theoretical Limits to Preserve n-D Distances in Lower Dimensions: Johnson-Lindenstrauss Lemma
3.3 Visual Knowledge Discovery Approaches
3.4 Point-to-Point Projections
3.5 General Line Coordinates to Convert n-D Points to Graphs
4 Case Studies on Visual Discovery
4.1 Case Study with GLC-L Algorithm
4.2 Case Study with the FSP Algorithm
4.3 Case Study with GLC-L+CNN Algorithm
4.4 Case Study with CPC-R+CNN Algorithm
4.5 Case Study with RPPR Algorithm
5 Scaling of Visual Discovery of ML Models
5.1 GLC and Embeddings to Cut Dimensions
5.2 Dimension Reduction and Visual PCA Interpretation with GLC
5.3 Cutting the Number of Points and Dealing with Complex Data
6 Visual Explanation of Analytical ML Models
6.1 Types of Explanations
6.2 Heatmap Pixel-Based Implicit Explanations
7 Conclusion and Future Work
References
Visual Analytics and Human Involvement in Machine Learning
1 Introduction
2 Overview: Visualizations Used During the Steps of the Machine Learning Process
3 Visualizations in the Steps of the ML Process
3.1 Data Preparation
3.2 Choosing a Model
3.2.1 Supervised Models
3.2.2 Unsupervised Models
3.2.3 Advanced Methods
3.3 Training the Model
3.4 Evaluation and Interpretation
3.4.1 Evaluation
3.4.2 Interpretation
3.4.3 Hyperparameters Tuning
3.4.4 Summary
4 User Interfaces and Frameworks
5 Discussion
References
Explainable Artificial Intelligence (XAI): Motivation, Terminology, and Taxonomy
1 Motivation
2 Explainability, Interpretability, and Related XAI Terms
3 Accuracy and Explainability
4 Segmentation of XAI Approaches
4.1 Complexity-Related Methods
4.2 Global and Local Interpretability Approaches
4.3 Model-Related Methods
4.3.1 Visualization
4.3.2 Knowledge Extraction
4.3.3 Influence Methods
4.3.4 Example-Based Explanation
5 Final Remark
References