An Efficient Svm and Aco-Rf Method for the Cluster-Based Feature Selection and Classification

Abstract

An efficient support vector machine (SVM) and ant colony optimization-random forest (ACO-RF) method for the cluster-based feature selection and classification were presented and evaluated. It has been evaluated on the breast cancer Wisconsin dataset. The complete process includes preprocessing, feature selection based on class levels clusters and classification. SVM has been applied first for the data classification. It has been applied based on the class labels and further associate the correlations based on the heatmap. Finally, ACO-RF based feature selection mechanism was applied. It will be helpful in the elimination of irrelevant features based on the threshold parameters. The result clearly indicates that SVM and ACO-RF outperforms in all aspects. It is also clear from the heatmap regarding the attribute correlation in all aspects.