A kernel-ensemble bagging support vector machine

Abstract

This paper proposes a kernel-ensemble bagging SVM classifier for binary class classification. The classifier is advantageous over bagging SVM classifiers because it has a two-phase grid search module, a proposed parameter randomization module and a proposed ranking module. The novel modules enhance the diversity thus improve the performance of the proposed SVM classifier. Six UCI datasets are used to evaluate the proposed kernel-ensemble bagging SVM. The result show that the proposed SVM classifier outperforms the single kernel bagging SVM classifiers.