DMCA: A Combined Data Mining Technique for Improving the Microarray Data Classification Accuracy

Data mining techniques are important to sift through the huge amount of gene expression values in microarrays resulting in valuable biological knowledge. An important example is classifying cancer samples, which is crucial to biologists for cancer diagnosis and treatment. In this paper we propose the DMCA technique in which the main objective is reducing the number of genes needed for accurate classification. The proposed technique is a combination of two feature selection techniques, f-score and entropy-based, and a powerful classifier, Support Vector Machines. DMCA achieved promising results and is characterized by being flexible in all of its stages. When applied to two public microarray datasets, DMCA succeeded in reducing the number of gene expression values needed to classify a sample by 71.29% and guaranteed reliable classification accuracy.