Improving support vector machine classifier by combining it with k nearest neighbor principle based on the best distance measurement

Li Rong et al. proposed a new method of improving the classification accuracy of support vector machine by combining it with k nearest neighbor principle, and presented two corresponding algorithms named KSVM and MPKSVM. Analyzing the actual reason why SVM classifies some test samples in error, this article points the deficiencies in the two algorithms. We bring forth an ameliorated method of improving generalization by combining SVM with the k nearest neighbor principle based on the best distance measurement, and give an algorithm named BSKSVM. The experimental results show that our algorithm can truly improve the classification accuracy.