Multi-label sampling based on local label imbalance

作者：

Highlights：

• The local imbalance is more crucial than the global one in multi-label data.

• The local imbalance based measure assesses the hardness of multi-label data.

• MLSOL and MLUL tackle the multi-label class imbalance issue via local imbalance.

• Suitable application situations of our two methods are identified, respectively.

摘要

•The local imbalance is more crucial than the global one in multi-label data.•The local imbalance based measure assesses the hardness of multi-label data.•MLSOL and MLUL tackle the multi-label class imbalance issue via local imbalance.•Suitable application situations of our two methods are identified, respectively.

论文关键词：Multi-label learning,Class imbalance,Oversampling and undersampling,Local label imbalance,Ensemble methods

论文评审过程：Received 30 December 2020, Revised 30 August 2021, Accepted 31 August 2021, Available online 2 September 2021, Version of Record 9 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108294