Angle Decrement Based Gaussian Kernel Width Generator for Support Vector Clustering

A new method to generate Gaussian kernel width parameter (q) for Support Vector Clustering (SVC) is proposed in this study. The proposed method is based on idea of decreasing angle, along with increment of q. This method is a modification of secant method that previously proposed. Experiments are performed using four sets of data, each data set has its own characteristics. Experimental results show that angle decrement based method can generates a valid sequence of q value with simpler computation than secant method. In general, angle decrement based method can improve the performance of SVC so that clustering process can be performed faster.