A comparative study of gene selection methods for cancer classification using microarray data

Abstract

Due to the high dimensionality of gene expression data, gene selection is an important step for improving gene expression data classification performance. This is true for the case of cancer classification using gene expression data. In this paper, we compare various feature selection methods that select appropriate number of genes as the features which are used for cancer classification. We have used several machine learning algorithms along with the different feature selection (gene) methods for developing a system for more accurately classifying cancer using microarray data. To prove effectiveness of the different gene selection methods, we have conducted a number of experiments that compare the cancer classification performance with and without performing gene selection. Results reveal that the classification system that performs gene selection obtains the better classification accuracy with a small number of genes.