Go to  Advanced Search

Genetic algorithm for feature selection and weighting for off-line character recognition

Show full item record

Files in this item

Files Size Format Description   View
ubc_2002-0117.pdf 5.722Mb Adobe Portable Document Format   View/Open
 
Title: Genetic algorithm for feature selection and weighting for off-line character recognition
Author: Hussein, Faten T.
Degree Master of Applied Science - MASc
Program Electrical and Computer Engineering
Copyright Date: 2002
Abstract: Computer-based pattern recognition is a process that involves several sub-processes, including pre-processing, feature extraction, classification, and post-processing. This thesis is involved with feature selection and feature weighting processes. Feature extraction is the measurement of certain attributes of the target pattern. Classification utilizes the values of these attributes to assign a class to the input pattern. In our view, the selection and weighting of the right set of features is the hardest part of building a pattern recognition system. The ultimate aim of our research work is the automation of the process of feature selection and weighting, within the context of character/symbol recognition systems. Our chosen optimization method for feature selection and weighting is the genetic algorithm approach. Feature weighting is the general case of feature selection, and hence it is expected to perform better than or at least the same as feature selection. The initial purpose of this study was to test the validity of this hypothesis within the context of character recognition systems and using genetic algorithms. However, our study shows that this is not true. We carried two sets of experimental studies. The first set compares the performance of Genetic Algorithm (GA)-based feature selection to GA-based feature weighting, under various circumstances. The second set of studies evaluates the performance of the better method (which turned out to be feature selection) in terms of optimal performance and time. The results of these studies also show that (a) in the presence of redundant or irrelevant features, feature set selection prior to classification is important for k-nearest neighbor classifiers; and (b) that GA is an effective method for feature selection and the performance obtained using genetic algorithms was comparable to that of exhaustive search. However, the scalability of GA to highly dimensional problems, although far superior to that of exhaustive search, is still an open problem.
URI: http://hdl.handle.net/2429/12103
Series/Report no. UBC Retrospective Theses Digitization Project [http://www.library.ubc.ca/archives/retro_theses/]

This item appears in the following Collection(s)

Show full item record

All items in cIRcle are protected by copyright, with all rights reserved.

UBC Library
1961 East Mall
Vancouver, B.C.
Canada V6T 1Z1
Tel: 604-822-6375
Fax: 604-822-3893