Semi-supervised time series classification on positive and unlabeled problems using cross-recurrence quantification analysis

作者：

Highlights：

• We show time-domain similarity measurements lead to inconsistent classification due to the noise and local differences;

• We use CRQA to compare time series recurrences on Positive and Unlabeled scenarios;

• Our approach has achieved better classification performances while classifying time series from natural phenomena.

摘要

•We show time-domain similarity measurements lead to inconsistent classification due to the noise and local differences;•We use CRQA to compare time series recurrences on Positive and Unlabeled scenarios;•Our approach has achieved better classification performances while classifying time series from natural phenomena.

论文关键词：Time series,Semi-supervised classification,Positive and unlabeled,Self-training,Phase space,Cross-recurrence quantification analysis

论文评审过程：Received 19 June 2017, Revised 9 February 2018, Accepted 25 February 2018, Available online 2 March 2018, Version of Record 11 March 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.02.030