ABSTRACT: The effectiveness of kernelized support vector machine in classification depends on the choice of kernel function, kernel parameter and soft margin parameter. In practice, there is need for proper guidance on the combination of kernel functions and soft margin parameters to be used. An insight into this is provided in this study. In this paper, we explore the notion of support vector machine and its kernelized version, investigate the performance of some kernel functions and soft margin parameters in support vector classification for some training sample sizes in Rd. We also examine the performance of kernelized support vector machine in functional setting and compare the classifier with maximum functional depth classification methods and centroid classifier based on simulation.
Keywords: Support vection machine, kernel functions, soft margin parameters, classification, error
JoST. 2018. 9(2): 21-28.
Accepted for Publication, March 23, 2018