A Hybrid Approach for Gene Selection and Classification using Support Vector Machine

Abstract

Deoxyribo Nucleic Acid (DNA) microarray technology allows us to generate thousands of gene expression in a single chip. Analyzing gene expression data plays vital role in understanding diseases and discovering medicines. Classification of cancer based on gene expression data is a promising research area in the field of bioinformatics and data mining. All genes do not contribute for efficient classification of samples. Hence, a robust feature selection method is required to identify the relevant genes which help in the classification of samples effectively. Most of the existing feature selection methods are computationally expensive. Redundancy in gene expression data leads to poor classification accuracy and also acts bad on multi class classification. This paper proposes an ensemble feature selection technique which is a combination of Recursive Feature Elimination (RFE) and Based Bayes error Filter (BBF) for gene selection and Support Vector Machine (SVM) algorithm for classification. The proposed ensemble gene selection method yields comparable performance on classification when compared to existing classifiers and provides a new insight in feature selection.