Mutual information based feature subset selection in multivariate time series classification

作者：

Highlights：

• A Feature subset selection pre-process is proposed for MTS classification.

• The proposed FSS improves accuracy and reduces the number of series.

• The selection procedure is based on the use of the temporal information of the series.

• High percentage of improvement is achieved for the dependent DTW dissimilarity based 1-NN classification.

摘要

•A Feature subset selection pre-process is proposed for MTS classification.•The proposed FSS improves accuracy and reduces the number of series.•The selection procedure is based on the use of the temporal information of the series.•High percentage of improvement is achieved for the dependent DTW dissimilarity based 1-NN classification.

论文关键词：Multivariate time series,Supervised classification,Feature susbset selection,Mutual information

论文评审过程：Received 24 October 2019, Revised 16 March 2020, Accepted 30 June 2020, Available online 10 July 2020, Version of Record 16 July 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107525