Robust Multi-combination Feature Selection for Microarray Data

Feature selection is an important used technique in data preprocessing for performing data mining on high-dimensional data sets. Feature selection often aims to select the best features to build a pattern classifier with reduced complexity, so as to achieve improved classification performance. Several feature selection algorithms are combined in order to produce more robust feature subsets and better classification results. In this paper, we propose a technique to combine the results of three different features selection algorithms to find the best subset. Experimental studies on gene-expression data set Leukemia show that our algorithm outperforms the three commonly used benchmark feature selection algorithm (RliefF, Information Gain, MRMR).