Over the past few years, has been shown that generalization power of Support Vector Machines (SVM) falls dramatically on imbalanced data-sets. In this paper, we propose a new method to improve accuracy of SVM on imbalanced ...