Revealing the structure of prediction models through feature interaction detection

作者：

Highlights：

• A method is proposed to detect feature interactions from prediction models.

• The method can help gain insight into the structure of prediction models.

• The method is based on high dimensional model representation.

• The method can detect any k-way interaction.

• The method can be used for both classification and regression models.

摘要

•A method is proposed to detect feature interactions from prediction models.•The method can help gain insight into the structure of prediction models.•The method is based on high dimensional model representation.•The method can detect any k-way interaction.•The method can be used for both classification and regression models.

论文关键词：Feature interaction,High dimensional model representation,Model interpretation,Black box

论文评审过程：Received 7 May 2021, Revised 9 November 2021, Accepted 9 November 2021, Available online 20 November 2021, Version of Record 29 November 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107737