A NOVEL HYBRID METHOD FOR GENE SELECTION IN MICROARRAY BASED CANCER CLASSIFICATION

摘要

Microarray technology allows researchers to measure the expression levels of many thousands of genes in a single experiment. One of the main applications of the microarray technology is cancer classification in which the problem associated with the analysis is to deal with high number of gene expression data that makes the processing more complex and time consuming. This paper aims to develop a classification algorithm by employing a hybrid method for gene selection. The hybrid method uses Analysis of variance (ANOVA) statistical technique and Principal Component Analysis (PCA) method that overcomes the problem by reducing the gene expression data in to minimal number of gene subsets (Differentially Expressed genes). Then k-nearest neighbors (KNN) classifier is applied on those gene subsets for cancer classification. Three publically available datasets namely leukemia, non-small lung and ovarian cancer data sets are used for classification. Experimental results shows that, the cosine and pearsons correlation distance function with ties rule provide higher classification accuracy and less misclassification error for those differentially expressed genes using different validation methods in KNN classifier.