Optimal Combination of SVM and Bayesian Density Model Using Dempster-Shafer Theory

In pattern classification, the diversity among classifiers is known to be able to provide complementary knowledge and improve classification performance, if properly exploited. In this paper, we propose to integrate the support vector machine (SVM) and the Bayesian density model by utilizing their respective posterior classification probabilities. The Dempster-Shafer (DS) theory was adopted for fusing the two classifiers. The effectiveness of the proposed method was verified using three datasets. The performance of the proposed approach was shown to be superior to that of the benchmark methods.