A Neighborhood Undersampling Stacked Ensemble (NUS-SE) in imbalanced classification

作者：

Highlights：

• An undersampling based stacked ensemble is proposed in imbalanced data learning.

• The proposed method tackles difficulties — class-imbalance, overlapping, and noise.

• It enhances classifier-diversity and sample-diversity in base learning.

• It uses an unbiased metadata with full utilization of training data.

• The proposed ensemble outperforms the existing stacked ensemble methods.

摘要

•An undersampling based stacked ensemble is proposed in imbalanced data learning.•The proposed method tackles difficulties — class-imbalance, overlapping, and noise.•It enhances classifier-diversity and sample-diversity in base learning.•It uses an unbiased metadata with full utilization of training data.•The proposed ensemble outperforms the existing stacked ensemble methods.

论文关键词：Imbalanced classification,Class imbalance,Stacked generalization,Stacking,Super learning,Stacked ensemble

论文评审过程：Received 30 July 2020, Revised 16 October 2020, Accepted 4 November 2020, Available online 7 November 2020, Version of Record 24 January 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.114246