A Multiple Spheres Weighted Support Vector Clustering Algorithm

Support vector clustering originates from the concept of support vector machine. It is a robust and unsupervised method. This paper improves the traditional SVC algorithm and presents a multiple spheres weighted support vector clustering (MW-SVC) algorithm. The MW-SVC algorithm is suitable for the normal condition that the clustering results of training samples are several clusters. It is fully automatic adaptively, runs fast and has highly accuracy rate. The results of experiment demonstrate the effectiveness of the proposed MW-SVC algorithm.