Abstract: Resampling technology has gradually become an important direction to solve the problem of classification for imbalanced data because of its simplicity and intuition.However,in the case of small data sets,under-sampling in resampling technology may lose important information of data sets,so oversampling is the focus of classification for imba-lanced data.Although the existing oversampling methods effectively overcome the imbalance between classes,they may cause dense areas of minority class to be denser,even lead to overlapping of samples.In addition,due to the noise of minority class,the existing oversampling methods may generate new samples around the noise,which makes the distribution of minority class more confusing.Aiming at these problems,this paper proposed a bi-directional oversampling method based on sample stratification.It firstly divides the minority samples into dense area and sparse area based on the highest density point and the intra-class average distance.And then the bi-directional oversampling is performed in the boundary region of dense area and the sparse area.In order to verify the effectiveness of the proposed algorithm,comprehensive experiments were conducted on 9 data sets of UCI database.The experimental results and Friedman test results show the superiority of the proposed algorithm for the task of imbalanced data classification.

Key words: Bi-directional oversampling, Classification, Dense area, Imbalanced data, Sparse area