K-nearest neighbor based bagging SVM pruning

A k-nearest neighbor (kNN) based bagging pruning algorithm for ensemble SVM classification is proposed in this paper. Redundant bags are discarded without reducing the performance of the ensemble SVM classifier. Ten VCI binary classification datasets are used to evaluate the performance of the proposed pruning algorithm against single SVM and bagging SVM classifiers. Results show that the proposed bagging SVM pruning improves the classification accuracies on most of the datasets with use less number of base classifiers thereby reducing computational requirements.