Optimizing support vector machine parameters using particle swarm optimization

Abstract

Support vector machine (SVM) is a popular classification method which is known to have a robust generalization capability. SVM calculates the best linear separator on the input feature space according to the training data. To classify data which are non-linearly separable, SVM uses kernel tricks to transform the data into a linearly separable data on a higher dimension feature space. Kernel trick uses various kinds of kernel functions, such as: linear kernel, polynomial, radial basis function (RBF) and sigmoid. Each function has parameters which affect the accuracy of SVM classification. In this research, we propose a combination of the SVM algorithm and particle swarm optimization (PSO) to search for the kernel parameters with the best accuracy. We use data from UCI repository of machine learning database: Diabetes, Image Letter Recognition and Yeast. The results indicate that the combination of SVM and PSO is effective in improving the accuracy of classification. PSO has been shown to be effective in systematically finding optimal kernel parameters for SVM, as a replacement for randomly choosing kernel parameters. Best accuracy for each data has been improved from previous research: 78.26% for Diabetes, 97.07% for Letter and 58.21% for Yeast. However, for bigger data sizes, this method may become impractical due to the high time requirement.