Feature selection for cancer classification contains a novel approach for feature selection for cancer microarray data using signal-to-noise ratio approach and t-statistics. It starts with a through overview of the concepts of gene expression data and feature selection approaches for cancer data sets. It then connects these concepts and applies them to the study of various literature and list out the approaches used and their limitations and advantages. Key features include; 1. A brief introduction on microarray data 2. Different feature selection approaches available in the literature are described 3. Provides proposed feature selection approach 4. Experimental evaluation and result analysis for different cancer data sets after classification.