Constructing Multiple Support Vector Machines Ensemble Based on Fuzzy Integral and Rough Reducts

Abstract

Even the multiple support vector machine (SVM) ensemble has been proved to improve the classification performance greatly than a single SVM, the classification result of the practically implemented SVM is often far from the theoretically expected level. As compared to traditional bagging and boosting methods, this paper proposes a novel SVM ensemble method based on fuzzy integral and rough reducts. In general, the proposed method is built in 3 steps: construct the individual SVM of ensemble by rough reduction technique; obtain the probabilistic outputs model of each component SVM; combine the component predictions based on fuzzy integral. The trained individual SVMs are aggregated to make a final decision. The simulating results demonstrate that the proposed multiple SVM ensemble method outperforms a single SVM and traditional SVM ensemble technique via bagging and boosting in terms of classification accuracy.