A novel approach to clustering analysis based on support vector machine

作者：Li ZH^*; Chen SB; Zheng RR; Wu JP; Mao ZY

来源：ADVANCES IN NEURAL NETWORKS - ISNN 2004, PT 1, SPRINGER-VERLAG BERLIN, HEIDELBERGER PLATZ 3, D-14197 BERLIN, GERMANY, 565-570, 2004.

This paper proposed a novel approach to clustering analysis for a large-scale data based on support vector machines (SVM). For conventional support vector clustering (SVC), data points are mapped by using a Gaussian kernel function to a high dimensional feature space. When mapped back to data space, this sphere is able to separate into several components. However, the dimension of feature space would be very high if a number of data points are mapped. This impairs the efficiency of SVM, and increases its computation. The approach in this paper utilized AIS to compress original data, and new reduced data points are obtained as the input of conventional SVC. Given elevator traffic data, simulation results indicated the applicability of this approach.

• 出版日期2004
• 单位华南理工大学