AdaptativeCC4.5: Credal C4.5 with a rough class noise estimator

作者：

Highlights：

• CC4.5 is a parametric classifier based on imprecise probabilities outperforming C4.5.

• That improvement increases in situations with class noise.

• Higher values of the parameter of CC4.5 are related with higher levels of noise.

• A new procedure to estimate the level of noise is combined with CC4.5.

• The new classifier is similar to the CC4.5 when its best parameter value is used.

摘要

•CC4.5 is a parametric classifier based on imprecise probabilities outperforming C4.5.•That improvement increases in situations with class noise.•Higher values of the parameter of CC4.5 are related with higher levels of noise.•A new procedure to estimate the level of noise is combined with CC4.5.•The new classifier is similar to the CC4.5 when its best parameter value is used.

论文关键词：Decision support system,Classification,Decision tree,Imprecise probabilities,Uncertainty measures,Noisy data

论文评审过程：Received 5 June 2017, Revised 7 August 2017, Accepted 26 September 2017, Available online 28 September 2017, Version of Record 5 October 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.09.057