Class-imbalanced positive instances augmentation via three-line hybrid

作者：

Highlights：

• A new heuristic imbalanced data positive instance augmentation technique THPIA is proposed.

• THPIA can effectively use the information of negative instances to synthesize representative pseudo-positive instances.

• THPIA further boosts the diversity of synthetic pseudo-positive instances through hybrid selection and uses distance constraints to prevent synthesizing noisy pseudo-positive instances in negative domains.

摘要

•A new heuristic imbalanced data positive instance augmentation technique THPIA is proposed.•THPIA can effectively use the information of negative instances to synthesize representative pseudo-positive instances.•THPIA further boosts the diversity of synthetic pseudo-positive instances through hybrid selection and uses distance constraints to prevent synthesizing noisy pseudo-positive instances in negative domains.

论文关键词：Class-imbalanced learning,Positive instances augmentation,Three-line hybrid,Oversampling,Classification

论文评审过程：Received 8 March 2022, Revised 13 September 2022, Accepted 13 September 2022, Available online 20 September 2022, Version of Record 30 September 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109902