Data Mining

Preface xiii
 
Preface to the Second Edition xv
 
Preface to the First Edition xvii
 
1 Data-Mining Concepts 1
 
1.1 Introduction 2
 
1.2 Data-Mining Roots 4
 
1.3 Data-Mining Process 6
 
1.4 From Data Collection to Data Preprocessing 10
 
1.5 Data Warehouses for Data Mining 15
 
1.6 From Big Data to Data Science 18
 
1.7 Business Aspects of Data Mining: Why a Data-Mining Project Fails? 22
 
1.8 Organization of This Book 26
 
1.9 Review Questions and Problems 28
 
1.10 References for Further Study 30
 
2 Preparing the Data 33
 
2.1 Representation of Raw Data 34
 
2.2 Characteristics of Raw Data 38
 
2.3 Transformation of Raw Data 40
 
2.4 Missing Data 43
 
2.5 Time-Dependent Data 44
 
2.6 Outlier Analysis 49
 
2.7 Review Questions and Problems 56
 
2.8 References for Further Study 59
 
3 Data Reduction 61
 
3.1 Dimensions of Large Data Sets 62
 
3.2 Features Reduction 64
 
3.3 Relief Algorithm 75
 
3.4 Entropy Measure for Ranking Features 77
 
3.5 Principal Component Analysis 80
 
3.6 Value Reduction 83
 
3.7 Feature Discretization: ChiMerge Technique 86
 
3.8 Case Reduction 90
 
3.9 Review Questions and Problems 93
 
3.10 References for Further Study 95
 
4 Learning from Data 97
 
4.1 Learning Machine 99
 
4.2 Statistical Learning Theory 104
 
4.3 Types of Learning Methods 110
 
4.4 Common Learning Tasks 112
 
4.5 Support Vector Machines 117
 
4.6 Semi-Supervised Support Vector Machines (S3VM) 131
 
4.7 kNN: Nearest Neighbor Classifier 134
 
4.8 Model Selection vs. Generalization 138
 
4.9 Model Estimation 142
 
4.10 Imbalanced Data Classification 150
 
4.11 90% Accuracy ... Now What? 154
 
4.12 Review Questions and Problems 158
 
4.13

MEHMED KANTARDZIC, PHD, is a Professor in the Department of Computer Engineering and Computer Science (CECS) at the University of Louisville, and is Director of the Data Mining Lab and CECS Graduate Programs. He is a member of IEEE, ISCA, KAS, WSEAS, IEE, and SPIE.