HYBRIDIZED UNIVARIATE AND MULTIVARIATE FILTER BASED APPROACHES FOR GENE SELECTION

Abstract

In this paper, we present a hybridized univariate and multivariate approach for optimal gene selection. This method seeks to find the optimal subset of genes of interest known as informative genes and maximize the accuracy of the model with reduction percentage. The merit of this approach is analyzed by experimental results on six benchmark datasets such as Leukemia, Breast Cancer and Lung Cancer, Hepatitis, Lymphoma and Embryonal Tumor and results showed an impressive accuracy. The classification accuracy of original dataset and informative genes selected using proposed model has been compared based on Naïve Bayesian, Multi-layer Perceptron and Support Vector Machine classifiers. Finally, the performance of selected informative gene subsets have been measured using few probabilistic measures such as; Balanced Classification Rate (BCR), F-Measure, Jaccard Index (JI), Adjusted Random Index (ARI), Normalized Mutual Information (NMI), and Purity and the results obtained from these measures establishes our algorithm for selecting informative genes.