Selective SVM ensemble based on discretization method

To improve the classification performance, a novel SVM ensemble learning algorithm based on discretization method and selective ensemble approach is proposed in this paper. This algorithm uses the discretized data sets obtained by the rough sets and Boolean reasoning method to construct individual SVMs with good diversity, which can improve the performance of ensemble learning. After every SVM is trained separately, a selective approach to SVM ensemble is utilized to reduce the ensemble size and improve its performance. Experimental comparison of the proposed algorithm against the traditional ensemble learning methods such as Bagging and Adaboost shows that it leads to small ensembles with better performance.