The class imbalance problem: A systematic study

Abstract In machine learning problems, dierences in prior class probabilities|or class imbalances|have been reported to hinder the performance of some standard classi ers, such as decision trees. This paper presents a systematic study aimed at answering three dierent questions. First, we attempt to understand what the class imbalance problem is by establishing a relationship between concept complexity, size of the training set and class imbalance level. Second, we discuss several basic re-sampling or cost-modifying methods previously proposed to deal with class imbalances and compare their eectiveness. Finally, we investigate the assumption that the class imbalance problem does not only aect decision tree systems but also aects other classi cation systems such as Neural Networks and Support Vector Machines