Efficient Parameter Selection for Support Vector Machines

Abstract

The support vector machines (SVM) is a popular classification method. Many users may not well tune hyperparameters because this step is time-consuming. However, the performance of SVM relies on the values of hyperparameters. To get around the problem, users may resort to anecdotal methods or default values set by software developers, but these methods may compromise the performance of classification accuracy. We investigate the theory that justifies P-SVM for tuning (γ,C) P-SVM significantly improved accuracy for classifying the business intelligence data. Experiments of simulation and real datasets show that P-SVM reducescomputational time substantially without much loss in accuracy.