Abstract: This paper reviewed the present situation of mining data in unbalanced classes at home and abroad in recent years. Firstly, it analysed in-depth the existing problems and their resulting nature. hhen it in detail dealt with various state-of-the-art data mining techniques under the unbalanced learning scenario. Moreover, from the data-level and algorithm-level respectively it analysed and compared them omprehensively in accordance with essential difference. At the same time, the paper summaricd measure metrics evaluating performance of mining imbalance data sets. Also, the paper pointed out recent hot issues of theoretic studies and applications. Finally, the perspectives on future work were also discussed.

Key words: Machine learning, Imbalanced classification, Resampling, Cost sensitive learning