Midpoint Validation Method for Support Vector Machine with Margin Adjustment Technique

Abstract

In this paper, we propose a midpoint-validation method and margin adjustment technique which improves the generalization of support vector machine. Margin adjustment technique enables the nearly effect as soft margin support vector machine by adjusting parameter. The midpoint-validation method creates midpoint data, as well as a turning adjustment parameter of support vector machine using midpoint data and previous training data. We compare its performance with the support vector machine, soft margin support vector machine, multilayer perceptron, radial basis function neural network and also tested our proposed method on fifth benchmark problems. The results obtained from the simulation shows the effectiveness of the proposed method.