Evidential classifier for imprecise data based on belief functions

作者：

Highlights：

• An evidential classifier (EC) working with credal classification is proposed.

• EC can reduce the error rate thanks to the introduction of meta-class.

• Credal classification is fit for dealing with the imprecise (overlapped) data set.

• A particular two-step combination procedure is introduced in EC.

• The discounted technique is involved in the fusion process to get reasonable result.

摘要

•An evidential classifier (EC) working with credal classification is proposed.•EC can reduce the error rate thanks to the introduction of meta-class.•Credal classification is fit for dealing with the imprecise (overlapped) data set.•A particular two-step combination procedure is introduced in EC.•The discounted technique is involved in the fusion process to get reasonable result.

论文关键词：Belief functions,Evidence theory,K-nearest neighbors,Credal classification,Outlier detection

论文评审过程：Received 18 October 2012, Revised 30 July 2013, Accepted 2 August 2013, Available online 17 August 2013.

论文官网地址：https://doi.org/10.1016/j.knosys.2013.08.005