An approach for raising the accuracy of one-class classifiers

Abstract

The support vector data description (SVDD) is a method proposed to solve the problem of one-class classification. It models a hypersphere around the target set, and by the introduction of kernel functions, more flexible descriptions are obtained. In SVDD, the width parameter s and the penalty parameter c have to be given beforehand by the user. To automatically optimize the values for these parameters, the error on both the target and outlier data has to be estimated. Because no outlier examples are available, we propose a max-min range method for generating artificial outliers in this paper. By generating artificial outliers around the target set, the accuracy of classifiers will improve. At the last, we use four benchmark data sets: Iris, Wine, Balance-scale, and Ionosphere data base to validate the approach in this research indeed has better classification result.