DESIGN OF ENSEMBLES WITH SVM KERNELS FOR DIABETES DATASET

Abstract

Data mining methods based on support vector machine are attractive to address the curse of dimensionality. The Kernel mapping contributes a unifying frame work for most of the commonly employed models to get the linear planes in the higher dimensional space. In this paper, we prove this approach enhances the accuracy of diabetes data set. We further refine the results with parameter tuning for the selected kernels. The natural question that arises in the case of many such different mappings to choose from, which is the best for a particular problem? The selection can be validated using independent test sets or a variety of data sets and methods of cross validations.