Online Multi-label Group Feature Selection

作者：

Highlights：

• Our work advances the relevance-, redundancy-, and interaction-based multi-label feature selection for managing streaming features.

• To the best of our knowledge, this is the first effort that accounts for intrinsic group structures of features and handling streaming features simultaneously in multi-label learning.

• OMGFS can be used to perform group- and single- feature selection, simultaneously.

• We validate the superiority of our proposed algorithm via comparing with other state-of-the-art multi-label feature selection algorithms from different performance views.

摘要

•Our work advances the relevance-, redundancy-, and interaction-based multi-label feature selection for managing streaming features.•To the best of our knowledge, this is the first effort that accounts for intrinsic group structures of features and handling streaming features simultaneously in multi-label learning.•OMGFS can be used to perform group- and single- feature selection, simultaneously.•We validate the superiority of our proposed algorithm via comparing with other state-of-the-art multi-label feature selection algorithms from different performance views.

论文关键词：Online feature selection,Multi-label learning,Streaming feature,Group feature selection

论文评审过程：Received 19 June 2017, Revised 6 December 2017, Accepted 7 December 2017, Available online 8 December 2017, Version of Record 3 February 2018.

论文官网地址：https://doi.org/10.1016/j.knosys.2017.12.008