Feature selection using data envelopment analysis

作者：

Highlights：

• Feature selection is regarded as a multi-index evaluation process.

• A novel feature selection framework based on data envelopment analysis is proposed.

• The framework evaluates features from a perspective of “efficient frontier” without parameter setting.

• A simple feature selection method is proposed based on super-efficiency DEA.

摘要

•Feature selection is regarded as a multi-index evaluation process.•A novel feature selection framework based on data envelopment analysis is proposed.•The framework evaluates features from a perspective of “efficient frontier” without parameter setting.•A simple feature selection method is proposed based on super-efficiency DEA.

论文关键词：Feature selection,Data envelopment analysis,Super-efficiency,Relevance,Redundancy

论文评审过程：Received 7 October 2013, Revised 8 March 2014, Accepted 25 March 2014, Available online 5 April 2014.

论文官网地址：https://doi.org/10.1016/j.knosys.2014.03.022