A ranking-based strategy to prune variable selection ensembles
作者：
Highlights：
• Ensemble pruning techniques are introduced in the context of variable selection.
• A ranking-based strategy is devised to prune a variable selection ensemble.
• The ensemble members are sorted by the prediction error of their associated models.
• Higher selection accuracy is gained by fusing fewer members ranked ahead.
• The superiority of the novel method over some other methods is validated.
摘要
•Ensemble pruning techniques are introduced in the context of variable selection.•A ranking-based strategy is devised to prune a variable selection ensemble.•The ensemble members are sorted by the prediction error of their associated models.•Higher selection accuracy is gained by fusing fewer members ranked ahead.•The superiority of the novel method over some other methods is validated.
论文关键词：Variable selection ensemble,Ensemble pruning,Selection accuracy,Aggregation order,Variable ranking,Stochastic stepwise selection
论文评审过程：Received 7 September 2016, Revised 24 January 2017, Accepted 29 March 2017, Available online 7 April 2017, Version of Record 21 April 2017.
论文官网地址：https://doi.org/10.1016/j.knosys.2017.03.031
