Training Data Selection for Support Vector Machines Model

In recent years, Support Vector Machines (SVM) have become a popular tool for pattern recognition and machine learning. In training a support vector machine we need to select the parameters of the kernel function as well as soft margin parameter C of SVM. Thus to develop the optimal classifier we need to determine the optimal kernel parameter and the optimal value of C. Determining the optimal classifier is called model selection. When applied to a large data set, however, it requires a long time for training so the model selection task and its performance can be degraded a long time. To speed up training thereby shortening the time for model selection, several methods have been proposed, one of which is to reduce the training set size called training data selection. Recently, there has been considerable research on data selection for SVM training. The main idea is to select only the patterns that are likely to be located near the decision boundary. In this paper we propose a methods that select a subset of data for SVM training. Our experimental results show that a significant amount of training data can be removed by our proposed method without degrading the performance of the resulting SVM classifiers.