FEATURE SELECTION AND CLASSIFICATION FOR MICROARRAY DATA USING UPGRADE CHI-SQUARE TEST

Abstract

Microarray technology allows the monitoring of thousands of gene expressions in various biological contexts.The purpose of this paper is to lay the groundwork for the proposal and creation of a new algorithm based on unbalanced classes that will upgrade the original Chi-square algorithm.The proposed UpgCHI over the Apache Spark framework based on unbalanced classes likely represents their efforts to provide a more robust and reliable solution in comparison to the original Chi-square method to handle multi-class problems.The results are compared with the original Chi-square test and an upgraded UpgCHI multiclass selection algorithm on microarray datasets using different three classifiers to evaluate the performance.The presented method generates good classification performance with an UpgCHI test, which demonstrates an increased accuracy of 96% in DLBCL 2Classes, 89% in Colon Tumor 2Classes, 95% in Leukemia 3Classes, 86% in Leukemia 4Classes and 100% in Brain Tumor 5Classes.