A Multi-class Classification Algorithm Based on Geometric Support Vector Machine.

A multi-class classification algorithm based on geometric support vector machine (SVM) is proposed. For each class of training samples, a convex hull is constructed in the sample space using the Schlesinger-Kozinec (SK) algorithm. For a sample to be classified, the class label is determined according to the convex hull in which it is located. If this sample is in more than one convex hull, or is not in any convex hull, the nearest neighbor rule is further employed. Subsequently, its class label is identified by the class of centroid closest to the sample. The experimental results show that compared with the existing multi-class SVM methods, the proposed algorithm can improve the classification accuracy.