Research on Cost-sensitive Classification Methods for Imbalanced Data

This paper is concerned with the class imbalance problem which has been known to hinder the learning performance of classification algorithms. The problem occurs when there are significantly more training samples available for some classes than for others. Various real-world datasets suffer from this phenomenon. The cost-sensitive classification algorithm has shown good classification performance on imbalanced datasets. Besides, active learning has shown potential for solving the class imbalance problem by providing the classifier more balanced datasets. In this paper, we propose a novel hybrid algorithm combining active learning and cost sensitive classification within the Metacost framework. Based on the selection strategies in active learning, our method can construct more informative balanced datasets for each base classifier in the Metacost framework, which can improve the classification performance. Our experimental results show that the proposed algorithm can achieve more competitive prediction performance in terms of Recall, G_mean and Area Under roc Curve.