Orthogonal variance decomposition based feature selection

作者：

Highlights：

• Efficient and technically sound feature selection based on Sobol decomposition.

• The proposed method takes into account interactions between features.

• Superior performance vis-a-vis other state of the art sampling techniques.

• The proposed method is built on solid mathematical framework.

摘要

•Efficient and technically sound feature selection based on Sobol decomposition.•The proposed method takes into account interactions between features.•Superior performance vis-a-vis other state of the art sampling techniques.•The proposed method is built on solid mathematical framework.

论文关键词：Feature selection,Variance decomposition,Sobol decomposition,Sensitivity index,Total sensitivity index,Wrapper methods,Data mining

论文评审过程：Received 11 April 2020, Revised 19 April 2021, Accepted 9 May 2021, Available online 18 May 2021, Version of Record 27 May 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115191