Pawel Cichosz, Department of Electronics and Information Technology, Warsaw University of Technology, Poland.

Acknowledgements xix

Preface xxi

References xxxi

Part I Preliminaries 1

1 Tasks 3

1.1 Introduction 3

1.2 Inductive learning tasks 5

1.3 Classification 9

1.4 Regression 14

1.5 Clustering 16

1.6 Practical issues 19

1.7 Conclusion 20

1.8 Further readings 21

References 22

2 Basic statistics 23

2.1 Introduction 23

2.2 Notational conventions 24

2.3 Basic statistics as modeling 24

2.4 Distribution description 25

2.5 Relationship detection 47

2.6 Visualization 62

2.7 Conclusion 65

2.8 Further readings 66

References 67

Part II Classification 69

3 Decision trees 71

3.1 Introduction 71

3.2 Decision tree model 72

3.3 Growing 76

3.4 Pruning 90

3.5 Prediction 103

3.6 Weighted instances 105

3.7 Missing value handling 106

3.8 Conclusion 114

3.9 Further readings 114

References 116

4 Naïve Bayes classifier 118

4.1 Introduction 118

4.2 Bayes rule 118

4.3 Classification by Bayesian inference 120

4.4 Practical issues 125

4.5 Conclusion 131

4.6 Further readings 131

References 132

5 Linear classification 134

5.1 Introduction 134

5.2 Linear representation 136

5.3 Parameter estimation 145

5.4 Discrete attributes 154

5.5 Conclusion 155

5.6 Further readings 156

References 157

6 Misclassification costs 159

6.1 Introduction 159

6.2 Cost representation 161

6.3 Incorporating misclassification costs 164

6.4 Effects of cost incorporation 176

6.5 Experimental procedure 180

6.6 Conclusion 184

6.7 Further readings 185

References 187

7 Classification model evaluation 189

7.1 Introduction 189

7.2 Performance measures 190

7.3 Evaluation procedures 213

7.4 Conclusion 231

7.5 Further readings 232

References 233

Part III Regression 235

8 Linear regression 237

8.1 Introduction 237

8.2 Linear representation 238

8.3 Parameter estimation 242

8.4 Discrete attributes 250

8.5 Advantages of linear models 251

8.6 Beyond linearity 252

8.7 Conclusion 258

8.8 Further readings 258

References 259

9 Regression trees 261

9.1 Introduction 261

9.2 Regression tree model 262

9.3 Growing 263

9.4 Pruning 274

9.5 Prediction 277

9.6 Weighted instances 278

9.7 Missing value handling 279

9.8 Piecewise linear regression 284

9.9 Conclusion 292

9.10 Further readings 292

References 293

10 Regression model evaluation 295

10.1 Introduction 295

10.2 Performance measures 296

10.3 Evaluation procedures 303

10.4 Conclusion 309

10.5 Further readings 309

References 310

Part IV Clustering 311

11 (Dis)similarity measures 313

11.1 Introduction 313

11.2 Measuring dissimilarity and similarity 313

11.3 Difference-based dissimilarity 314

11.4 Correlation-based similarity 321

11.5 Missing attribute values 324

11.6 Conclusion 325

11.7 Further readings 325

References 326

12 k-Centers clustering 328

12.1 Introduction 328

12.2 Algorithm scheme 330

12.3 k-Means 334

12.4 Beyond means 338

12.5 Beyond (fixed) k 342

12.6 Explicit cluster modeling 343

12.7 Conclusion 345

12.8 Further readings 345

References 347

13 Hierarchical clustering 349

13.1 Introduction 349

13.2 Cluster hierarchies 351

13.3 Agglomerative clustering 353

13.4 Divisive clustering 361

13.5 Hierarchical clustering visualization 364

13.6 Hierarchical clustering prediction 366

13.7 Conclusion 369

13.8 Further readings 370

References 371

14 Clustering model evaluation 373

14.1 Introduction 373

14.2 Per-cluster quality measures 376

14.3 Overall quality measures 385

14.4 External quality measures 393

14.5 Using quality measures 397

14.6 Conclusion 398

14.7 Further readings 398

References 399

Part V Getting Better Models 401

15 Model ensembles 403

15.1 Introduction 403

15.2 Model committees 404

15.3 Base models 406

15.4 Model aggregation 420

15.5 Specific ensemble modeling algorithms 431

15.6 Quality of ensemble predictions 448

15.7 Conclusion 449

15.8 Further readings 450

References 451

16 Kernel methods 454

16.1 Introduction 454

16.2 Support vector machines 457

16.3 Support vector regression 473

16.4 Kernel trick 482

16.5 Kernel functions 484

16.6 Kernel prediction 487

16.7 Kernel-based algorithms 489

16.8 Conclusion 494

16.9 Further readings 495

References 496

17 Attribute transformation 498

17.1 Introduction 498

17.2 Attribute transformation task 499

17.3 Simple transformations 504

17.4 Multiclass encoding 510

17.5 Conclusion 521

17.6 Further readings 521

References 522

18 Discretization 524

18.1 Introduction 524

18.2 Discretization task 525

18.3 Unsupervised discretization 530

18.4 Supervised discretization 533

18.5 Effects of discretization 551

18.6 Conclusion 553

18.7 Further readings 553

References 556

19 Attribute selection 558

19.1 Introduction 558

19.2 Attribute selection task 559

19.3 Attribute subset search 562

19.4 Attribute selection filters 568

19.5 Attribute selection wrappers 588

19.6 Effects of attribute selection 593

19.7 Conclusion 598

19.8 Further readings 599

References 600

20 Case studies 602

20.1 Introduction 602

20.2 Census income 605

20.3 Communities and crime 631

20.4 Cover type 640

20.5 Conclusion 654

20.6 Further readings 655

References 655

Closing 657

A Notation 659

A.1 Attribute values 659

A.2 Data subsets 659

A.3 Probabilities 660

B R packages 661

B.1 CRAN packages 661

B.2 DMR packages 662

B.3 Installing packages 663

References 664

C Datasets 666

Index 667

The New copy of this book will include any supplemental materials advertised. Please check the title of the book to determine if it should include any access cards, study guides, lab manuals, CDs, etc.

The Used, Rental and eBook copies of this book are not guaranteed to include any supplemental materials. Typically, only the book itself is included. This is true even if the title states it includes any access cards, study guides, lab manuals, CDs, etc.