Abstract: To significantly improve the classification performance of the minority class,we present an over-sampling method based on feature selection.Firstly,feature selection is performed on the training data set in order to select a set of key columns.Then minority class samples are produced using selected key columns,and each sample consists of two kinds of features.One type of features is characteristic value that is corresponding to the selected key columns,the others is generated according to the principle of SMOTE.Comparing to SMOTE algorithm,results show that the new method performs better than SMOTE,and it can effectively reduce the imbalance of data and improve the classification accuracy of the minority class.