Several novel evaluation measures for rank-based ensemble pruning with applications to time series prediction

作者：

Highlights：

• Four evaluation measures for rank-based ensemble pruning for time series prediction are proposed.

• The proposed measure ReTSP-Trend takes into consideration the trend of time series.

• ReTSP-Trend guarantees the predictor supplementing the subensemble the most will be selected.

• ReTSP-Trend remarkably improves the predictive ability of the pruned ensembles.

摘要

•Four evaluation measures for rank-based ensemble pruning for time series prediction are proposed.•The proposed measure ReTSP-Trend takes into consideration the trend of time series.•ReTSP-Trend guarantees the predictor supplementing the subensemble the most will be selected.•ReTSP-Trend remarkably improves the predictive ability of the pruned ensembles.

论文关键词：Ensemble pruning,Time series prediction,Rank-based ensemble pruning,Complementarity measure for time series prediction (ComTSP),Concurrency thinning for time series prediction (ConTSP),Reduce Error pruning for time series prediction (ReTSP-Trend),Time window size

论文评审过程：Available online 7 August 2014.

论文官网地址：https://doi.org/10.1016/j.eswa.2014.07.049