Hybridizing mRMR and harmony search for gene selection and classification of microarray data

Gene selection, aimed at eliminating noisy and irrelevant genes, plays a crucial role in the analysis of microarray data. In this paper, we propose to hybridize a filter and a wrapper method in selecting discriminative genes for the classification of microarray data. First, the minimum Redundancy Maximum Relevance (mRMR) algorithm is exploited to select a subset of genes that are relevant to the disease and less redundant to each other from the original gene space. Then, harmony search algorithm combined with a classifier works on the reduced gene subset in a stochastic way to further explore the gene subset and obtain a more discriminative gene subset. To verify the effectiveness of the proposed approach, two other well-performed gene selection algorithms, ReliefF and FCBF, are involved as well, and experimental comparisons on six publicly available microarray data were conducted with 1-Nearest-Neighbor and Naive Bayes classifiers, respectively. Experimental results demonstrate that our approach greatly improves the classification accuracy and outperforms others for both two-category and multi-category problems. Copyright © 2015 Binary Information Press.