A mixed integer linear programming support vector machine for cost-effective feature selection
作者：
Highlights：
• To develop a cost-effective feature selection model that can achieve acceptable classification accuracy with a significantly reduced cost.
• To formulate a robust counterpart of the cost-effective feature selection model that can address the uncertainty of feature costs.
• To develop a heuristic algorithm to improve computational efficiency by tightening the bound of an integer budget constraint.
• To validate the effectiveness of the proposed models and heuristic algorithm on a variety of benchmark and synthetic datasets.
• To technically review recent cost-based feature selection methods and compare the performances with our proposed cost-effective model.
摘要
•To develop a cost-effective feature selection model that can achieve acceptable classification accuracy with a significantly reduced cost.•To formulate a robust counterpart of the cost-effective feature selection model that can address the uncertainty of feature costs.•To develop a heuristic algorithm to improve computational efficiency by tightening the bound of an integer budget constraint.•To validate the effectiveness of the proposed models and heuristic algorithm on a variety of benchmark and synthetic datasets.•To technically review recent cost-based feature selection methods and compare the performances with our proposed cost-effective model.
论文关键词：Feature selection,Support vector machine,Mixed integer linear programming,Robust optimization,Feature cost,Cost uncertainty
论文评审过程：Received 10 February 2020, Revised 5 May 2020, Accepted 13 June 2020, Available online 15 June 2020, Version of Record 22 June 2020.
论文官网地址：https://doi.org/10.1016/j.knosys.2020.106145
