A new density-based subspace selection method using mutual information for high dimensional outlier detection

作者：

Highlights：

• New measures to select relevant subspace for high dimensional outlier detection.

• A density-based representation for unsupervised subspace selection.

• Introducing “Maximum Relevance to Density, MRD” criterion to select subspace.

• Proposing “minimum Redundancy Maximum Relevancy to Density, mRMRD” criterion.

• The proposed method based on MRD/mRMRD statistically outperforms benchmark methods.

摘要

•New measures to select relevant subspace for high dimensional outlier detection.•A density-based representation for unsupervised subspace selection.•Introducing “Maximum Relevance to Density, MRD” criterion to select subspace.•Proposing “minimum Redundancy Maximum Relevancy to Density, mRMRD” criterion.•The proposed method based on MRD/mRMRD statistically outperforms benchmark methods.

论文关键词：High dimensional data,Relevant subspace selection,Density-based representation,Maximum-Relevance-to-Density,minimum-Redundancy-Maximum-Relevance-to-Density,Mutual information

论文评审过程：Received 10 October 2020, Revised 19 December 2020, Accepted 29 December 2020, Available online 2 January 2021, Version of Record 23 January 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2020.106733