Online multi-label streaming feature selection based on neighborhood rough set

作者：

Highlights：

• A new neighborhood relation is proposed to effectively solve the problem of granularity selection in neighborhood rough set.

• We generalize classical neighborhood rough set model to fit multi-label learning and present a novel measure to compute positive region.

• We propose a new feature selection framework, which solves online streaming feature selection and multi-label feature selection simultaneously.

• The experiment on ten benchmark datasets with different application scenarios shows a competitive performance of our proposed method against the state-of-the-art multi-label feature selection algorithms.

摘要

•A new neighborhood relation is proposed to effectively solve the problem of granularity selection in neighborhood rough set.•We generalize classical neighborhood rough set model to fit multi-label learning and present a novel measure to compute positive region.•We propose a new feature selection framework, which solves online streaming feature selection and multi-label feature selection simultaneously.•The experiment on ten benchmark datasets with different application scenarios shows a competitive performance of our proposed method against the state-of-the-art multi-label feature selection algorithms.

论文关键词：Online feature selection,Multi-label learning,Neighborhood rough set,Granularity

论文评审过程：Received 15 November 2017, Revised 18 June 2018, Accepted 16 July 2018, Available online 18 July 2018, Version of Record 25 July 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.07.021