Asymmetrical Support Vector Machine Based on Moving Optimal Separating Hyperplane

Abstract

Aiming at the problem about classifying two samples, support vector machine (SVM) put forward by Vapnik didn't think over the difference of two classes of classification error, so a new method, asymmetrical support vector machine (A-SVM), is given. The optimal separating hyperplane was deviated from the optimal support hyperplane of some kind of sample set by parallel moving the optimal separating hyperplane, and then this kind of sample set could be recognized with higher accurate ratio. Example result shows that A-SVM is similar to SVM for the total recognizing performance of both learning and testing. However, A-SVM is better than SVM when separating the kind of sample set