A New Over-sample Method Based on Distribution Density
Kuoyi Shao, Yun Zhai, Haifeng Sui, Changsheng Zhang, Nan Ma · Journal of Computers · 2014
A new method was proposed for leaning from the imbalanced dataset based the samples distribution density in this paper. In the proposed scheme, a model of samples distribution density was designed, followed by the improved smote progress SDD-SMOTE where we smoted the minority samples according to the samples distribution density. Cross-validation results show that proposed SDD-SMOTE method to some extent improves the minority prediction in both the recall and the precision metrics.