Dynamic relevance and interdependent feature selection for continuous data
作者：
Highlights：
• A novel Feature Selection (FS) named DRIFS is proposed based on statistical measures.
• The DRIFS eliminates discretization overhead for continuous data.
• The DRIFS is on a par with objectives of information theoretic based FS.
• The DRIFS method differentiates between redundant and interdependent features.
• Application of the DRIFS leads to an improvement in classification accuracy.
摘要
•A novel Feature Selection (FS) named DRIFS is proposed based on statistical measures.•The DRIFS eliminates discretization overhead for continuous data.•The DRIFS is on a par with objectives of information theoretic based FS.•The DRIFS method differentiates between redundant and interdependent features.•Application of the DRIFS leads to an improvement in classification accuracy.
论文关键词：Feature selection,Dynamic relevance,Interdependent features,Redundant features
论文评审过程：Received 21 March 2020, Revised 26 October 2021, Accepted 25 November 2021, Available online 4 December 2021, Version of Record 8 December 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.116302
