Learning Biased SVM with Weighted Within-Class Scatter for Imbalanced Classification

Abstract

Support vector machine (SVM) is a powerful tool for pattern classification and regression estimation. However, for the class imbalanced problem, conventional SVMs are not suitable to the imbalanced learning tasks since they tend to misclassify the minority class, which is always the more important class. In this paper, we propose an improved biased SVM with weighted within-class structure for imbalanced classification. The new algorithm makes the minority class more clustered by assigning a small weight for the within-class scatter matrix of minority class, which can improve the classification performance. The experimental results on several benchmark datasets demonstrate the effectiveness of the proposed algorithm for imbalanced data classification problems.