Support Vector Clustering Algorithm Based on WGKMW

ZHAO Li-feng,WANG Dong

DOI: https://doi.org/10.3969/j.issn.1004-373x.2011.10.024

2011-01-01

Abstract:In allusion to the problem that Gaussian radial basic function restraints the generalization performance of support vector machines in support vector clustering and influences the effect of clustering when data distribution isn't uniform,a support vector clustering algorithm based on weighted gaussian kernel with multiple widths is proposed.The new kernel has more adjustable parameters than Gaussian radial basic function for controlling the data to improve the generalization performance.Simulation results show that weighted Gaussian kernel with multiple widths can more effectively cluster on data,compared with Gaussian radial basic function.