A multiple-filter-GA-SVM method for dimension reduction and classification of DNA-microarry data

Abstract

The following article proposes a Multiple-Filter by using a genetic algorithm (GA) combined with a support vector machine (SVM) for gene selection and classification of DNA microarray data. The proposed method is designed to select a subset of relevant genes that classify the DNA-microarray data more accurately. First, three traditional statistical methods are used for gene selection. Then different relevant gene subsets are selected by using a GA/SVM framework using leave-one-out cross validation (LOOCV) to avoid data overfitting. A gene subset (niche), consisting of relevant genes, is obtained from each statistical method, by analyzing the frequency of each gene in the different gene subsets. Finally, the most frequent genes contained in the niche, are evaluated by the GA/SVM to obtain a final relevant gene subset. The proposed method is tested in two DNA-microarray datasets: Leukemia and colon. In the experimental results it is observed that the Multiple-Filter-GA-SVM (MF-GA-SVM) work very well by achieving lower classification error rates using a smaller number of selected genes than other methods reported in the literature.