Feature Selection for Cancer Classification Using Microarray Gene Expression Data

Abstract

Due to the high-dimension nature of microarray data and their small sample sizes, microarray data impose a great challenge to computational techniques.In order to tackle the difficulties in analyzing microarray data, the apparent need of feature selection methods was realized by researchers; see, e.g.Alon et al. [3], Golub et al. [4], Ross et al. [5] & Ben-Dor et al. [6] among many others.This has led to a recent surge of dimension reduction approaches presented in both bioinformatics and statistics.According to the literature, it is widely believed that in most microarray gene expression data, only some relevant genes play an important role in classification and the rest are irrelevant to