I=
Pakistan Research Repository Home
 

Title of Thesis

Optimization of Classifiers using Genetic Programming

Author(s)

Abdul Majid

Institute/University/Department Details
Faculty of Computer Science and Engineering / Ghulam Ishaq Khan Institute of Engineering Sciences and Technology, Topi
Session
2006
Subject
Engineering Sciences
Number of Pages
155
Keywords (Extracted from title, table of contents and abstract of thesis)
Classifiers, Genetic, Programming, Successfulness, Experimental, Heterogeneous, Promising, Homogenous, Optimization

Abstract
The success of pattern classification system depends on the improvement of its classification stage.The work of thesis has investigated the potential of Genetic Programming (GP) search space to optimize the performance of various classification models. In this thesis, two GP approaches are proposed. In the first approach, GP is used to optimize the performance of individual classifiers. The performance of linear classifiers and nearest neighbor classifiers is improved during GP evolution to develop a high performance numeric classifier. In second approach, component classifiers are trained on the input data and their predictions are extracted. GP search space is then used to combine the predictions of component classifiers to develop an optimal composite classifier (OCC). This composite classifier extracts useful information from its component classifiers during evolution process. In this way, the decision space of composite classifier is more informative and discriminant. Effectiveness of GP combination technique is investigated for four different types of classification models including linear classifiers, support vector machines (SVMs) classifiers, statistical classifiers and instance based nearest neighbor classifiers.
The successfulness of such composite classifiers is demonstrated by performing various experiments, while using Receiver Operating Characteristics (ROC) curve as the performance measure. It is evident from the experimental results that OCC outperforms its component classifiers. It attains high margin of improvement at small feature sets. Further, it is concluded that classification models developed by heterogeneous combination of classifiers have more promising results than their homogenous combination.
GP optimization technique automatically caters the selection of suitable component classifiers and model selection. Two main objectives are achieved, while using GP optimization. First, objective achieved is the development of more optimal classification models. The second one is the enhancement in the GP search strategy itself.

Download Full Thesis
2,739 KB
S. No. Chapter Title of the Chapters Page Size (KB)
1 0 CONTENTS

 

vi
120 KB
2

1

THEORETICAL ASPECTS

1.1 Introduction And Motivations
1.2 Definitions
1.3 Thesis Goals
1.4 Contributions Of Thesis
1.5 List Of Publications
1.6 Thesis Organization

1
230 KB
3 2 PATTERN CLASSIFICATION SYSTEM


2.1 Pattern Classification System
2.2 Designing A Classification System

11
543 KB
4 3 GENETIC PROGRAMMING IN CLASSIFICATION PROBLEMS

3.1 Introduction And Motivations
3.2 Machine Learning In Classification Problems
3.3 Gp Based Optimization Of Classifiers
3.4 GP Based Multi-objective Optimization
3.5 Advantages And Disadvantages Of GP Based Optimization Methods

30
367 KB
5 4 GENETIC PROGRAMMING FOR COMBINING CLASSIFIERS

4.1 Introduction And Motivations
4.2 Categorization Of Combination Of Classifiers
4.3 Issues In Combining Classifiers
4.4 GP In Combining Classifiers

55
277 KB
6 5 COMBINING LINEAR CLASSIFIERS

5.1 Introduction And Motivations
5.2 Chapter Goals
5.3 Linear Classifier
5.4 GP Based Combination Of Linear Classifiers

68
495 KB
7 6 COMBINING SUPPORT VECTOR MACHINES

6.1 Introduction And Motivations
6.2 Chapter Goals
6.3 SVM Classifiers
6.4 Proposed Methodology: Combining SVM Classifiers
6.5 Implementation Details
6.6 Results And Discussion
6.7 Summary Of The Chapter

77
558 KB
8 7 COMBINING STATISTICAL CLASSIFIERS

7.1 Introduction And Motivations
7.2 Chapter Goals
7.3 Analysis Of Classifiers
7.4 Combining Statistical Classifiers
7.5 Summary Of The Chapter

97
752 KB
9 8 OPTIMIZATION AND COMBINATION OF NEAREST NEIGHBOR CLASSIFIER

8.1 Introduction And Motivations
8.2 Chapter Goals
8.3 Experimental Setup
8.4 Developing Numeric Expression Classifier (modnn)
8.5 Results And Discussion
8.6 Combining Nn Classifiers
8.7 Summary Of The Chapter

108
345 KB
10 9 CONCLUSIONS

9.1 General Conclusions
9.2 Future Directions
 

121
209 KB
11 10 REFERENCES & APPENDICES

126


246 KB