An analysis of heuristic metrics for classifier ensemble pruning based on ordered aggregation

作者:

Highlights:

• An up-to-date revision and analysis of ordering-based ensemble pruning methods.

• Analysis of how the pruning accuracy is affected by the size of the original ensemble.

• The accuracy of pruned ensembles is shown to be superior to the stable predictions given by bagged ensembles.

• It is shown how the pruning metrics perform in binary and multiclass classification tasks.

• A thorough analysis of the prediction consistency, time and space complexities of pruning methods is performed.

摘要

•An up-to-date revision and analysis of ordering-based ensemble pruning methods.•Analysis of how the pruning accuracy is affected by the size of the original ensemble.•The accuracy of pruned ensembles is shown to be superior to the stable predictions given by bagged ensembles.•It is shown how the pruning metrics perform in binary and multiclass classification tasks.•A thorough analysis of the prediction consistency, time and space complexities of pruning methods is performed.

论文关键词:Heuristic optimization,Ensemble selection,Ensemble pruning,Classifier ensemble,Machine learning,Difficult samples,Ordering-based pruning,Classifier complementariness

论文评审过程:Received 10 September 2020, Revised 3 June 2021, Accepted 4 December 2021, Available online 7 December 2021, Version of Record 20 December 2021.

论文官网地址:https://doi.org/10.1016/j.patcog.2021.108493

Back to Top