SMOTE-NaN-DE: Addressing the noisy and borderline examples problem in imbalanced classification by natural neighbors and differential evolution

作者：

Highlights：

• A novel oversampling method based on natural neighbors and differential evolution.

• A novel error detection based on natural neighbors.

• The differential evolution is modified to optimize found noisy and borderline samples.

• The proposed algorithm is applied to 4 oversampling methods and outperforms 9 SMOTE-based methods.

摘要

•A novel oversampling method based on natural neighbors and differential evolution.•A novel error detection based on natural neighbors.•The differential evolution is modified to optimize found noisy and borderline samples.•The proposed algorithm is applied to 4 oversampling methods and outperforms 9 SMOTE-based methods.

论文关键词：Class-imbalance learning,Class-imbalance classification,Oversampling,Differential evolution,Natural neighbors

论文评审过程：Received 17 January 2021, Revised 14 April 2021, Accepted 15 April 2021, Available online 17 April 2021, Version of Record 23 April 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107056