Improving ADABoost Algorithm with Weighted SVM for Imbalanced Data Classification

Recently, different boosting algorithms have been proposed in order to improve the performance of classification for imbalanced data. In this paper, we present an improved ADABoost algorithm, called Im.ADABoost, for imbalanced data including two main improvements: (i) initializing different error weights adapted to the imbalance rate of the datasets; (ii) calculating the confidence weights of the member classifier that is sensitive to the total errors caused on the positive label. Additionally, we combine Im.ADABoost with Weighted-SVM to enhance classification efficiency on imbalanced datasets. Our experimental results show some promising potential of the proposed algorithm.