A Ga-Based Multivariate Feature Selection Algorithm For Cancer Identification.

SUBARI @ RAHMAT, NOOR FAUZANA (2010) A Ga-Based Multivariate Feature Selection Algorithm For Cancer Identification. Other thesis, Universiti Teknologi Malaysia.

[img]
Preview
PDF
noorfauzanaac060176d10ttp.PDF

Download (135Kb) | Preview
[img]
Preview
PDF
noorfauzanaac060176d10ttt.PDF

Download (1939Kb) | Preview

Abstract

This study is all about the analysis of cancer microarray dataset by doing feature selection and validating the results using unbiased protocol. The problem of cancer disease facing today make the urgent need of cancer identification towards gene expression of the patient’s sample got from microarray technology. The high-dimensional, noise and redundant of the microarray data have motivate previous researchers to propose some feature selection techniques to make the analysis. However, the analysis done before was in limited scope. Hence, the comparative analysis that can be make only in a small area. Besides, most of the analysis are not stated systematically the unbiased protocol that should be used to avoid bias results. Hence, this study proposed a new architecture framework which used Genetic Algorithm based feature selection techniques and classification methods of SVM and k-NN. Double loop cross-validation is used as the unbiased protocol to validate the results from classification. There are 3 cancer datasets will be undergo in this study. In the end of this study, it is found that the results can avoid the overoptimist by the usage of the unbiased protocol and GA-SVM with k-NN classifier is the best method for cancer identification.

Item Type: Thesis (Other)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Faculty of Computer Science and Information System > Information System
Depositing User: Unnamed user with email knizam@utm.my
Date Deposited: 04 Jul 2013 06:58
Last Modified: 04 Jul 2013 06:58
URI: http://ir.fsksm.utm.my/id/eprint/5

Actions (login required)

View Item View Item