New Support Vector Machine for Imbalance Data Classification

Support vector machine has a better classification and prediction performance on balance data classification, but has a poor performance for imbalance data. In this paper, the reason of the poor results which were produced by imbalance data is explained. And a new approach is proposed to solve the imbalance data classification. By this method the imbalance data problem is converted to several independent classifications with balance data and can be trained in parallel. The experiments show that new method has a best effect in some classifying strategy.