Strategies for Dealing with Real World Classification Problems

..................................................................................................................................... 4 1.1 Motivation ................................................................................................................... 5 1.2 Thesis Overview .......................................................................................................... 7 2 Data mining: Concepts and Definitions ........................................................................... 10 2.1 Data Mining as Process ............................................................................................. 11 2.2 Knowledge Extraction Tasks .................................................................................... 12 2.3 Classification Methods .............................................................................................. 13 2.4 Model Evaluation ...................................................................................................... 14 2.4.1 Evaluation Tactics .............................................................................................. 15 2.4.2 Metrics ............................................................................................................... 15 3 Handling Incomplete Data ............................................................................................... 18 3.1 Problem Statement .................................................................................................... 18 3.2 Methods for Dealing with Missing Data ................................................................... 19 3.2.1 Filter-based MDTs ............................................................................................. 19 3.2.2 Imputation-based MDTs .................................................................................... 20 3.2.3 Embedded MDTs ............................................................................................... 21 3.3 A New Method for Data Imputation ......................................................................... 22 3.3.1 Method Description ........................................................................................... 22 3.3.2 Experimental Evaluation .................................................................................... 23 3.4 Conclusions on Data Imputation ............................................................................... 24 4 Feature Selection .............................................................................................................. 27 4.1 Problem Statement .................................................................................................... 27 4.2 Feature Selection Techniques.................................................................................... 28 4.2.1 Search Strategies ................................................................................................ 29 4.2.2 Evaluation Measures .......................................................................................... 29 4.2.3 Filter Methods .................................................................................................... 31 4.2.4 Wrapper Methods............................................................................................... 32 4.3 Combining Generation Strategies ............................................................................. 33 4.4 Experimental Evaluation ........................................................................................... 34 4.4.1 Evaluating the Wrapper Methodology ............................................................... 34 4.4.2 Evaluating the Combination Strategy ................................................................ 37 4.5 Conclusions on Feature Selection ............................................................................. 39