Online local pool generation for dynamic classifier selection

作者：

Highlights：

• An online local pool generation method for Dynamic Classifier Selection is proposed.

• The method produces locally accurate classifiers for instances in overlap regions.

• Results show a competent classifier is more often selected when using the local pool.

• The proposed method also outperforms Bagging and a globally generated pool.

摘要

•An online local pool generation method for Dynamic Classifier Selection is proposed.•The method produces locally accurate classifiers for instances in overlap regions.•Results show a competent classifier is more often selected when using the local pool.•The proposed method also outperforms Bagging and a globally generated pool.

论文关键词：Multiple classifier systems,Instance hardness,Pool generation,Dynamic classifier selection

论文评审过程：Received 14 May 2018, Accepted 1 August 2018, Available online 10 August 2018, Version of Record 18 August 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.08.004