Multilevel Feature Selection Method for Improving Classification of Microarray Gene Expression Data

Microarray gene expression profiles provide valuable answers to a variety of problems, and contributes advances in clinical medicine. Gene data typically has high dimension small sample size. selection from microarray is challenge due dimensionality the data. The number samples dataset much smaller compared genes as features. To extract useful information cancer reduce dimensionality, significant necessary. An effective method feature helps reduction improves classification performance. Experimental results suggest that appropriate combination filter methods more than individual techniques for classification. In this paper, we propose two-layered method. first layer, t-test statistical used remove features have little correlation with results. second line segment approximation transform subset into less dimensional space. Four well known classifiers kNN, SVM, NBC, DT were verify performance proposed algorithm on binary class experimental show can effectively select relevant subsets, achieves higher accuracy.