No abstract

Abstract-Support Vector Machine (SVM) has proven track record in Classification. Higher level of accuracy and reasonably good speed factor is attracting the analysts towards SVM classification. The core of success of SVM accuracy is its Kernel. The right choice of kernel can lead to highest level of accuracy whereas if the choice is not good then the accuracy may be too low to be considerable. In this paper, methods to choose the right kernel are listed with comparative analysis of multiple kernels on same data set. Multiple data sets are implemented on various kernels to establish the impact of kernel on SVM classification.