Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Extended Compressed Random Shapelet Forest for Multivariate Time Series Classification
1Chengdu Computer Application Institute CAS: Chengdu Information Technology of Chinese Academy of Sciences Co. Ltd., China.
2Key Laboratory of Internet Natural Language Processing of Sichuan Provincial Education Department, Leshan Normal University, China /EADDRESS/3Sichuan Provincial Key Laboratory of Philosophy and Social Science for Language Intelligence in Special Education, Leshan Normal University, China /EADDRESS/4School of Computer Science and Technology, University of Chinese Academy of Sciences, China /EADDRESS/
ABSTRACT
Time series classification (TSC) has attracted considerable attention from the time series data mining community in the last decade. In our previous work, a Compressed Random Shapelet Forest (CRSF) has been proposed to handle this issue. The CRSF algorithm not only achieves competitive performance compared with the state-of-the-art algorithms of TSC but also provides good interpretability of classification results. However, it can only support the classification of univariate time series. This paper introduces a classification algorithm, ECRSF, which extends from the CRSF algorithm for multivariate time series classification. It employs SAX to compress the time series and the Shapelet, which is a pattern that maximally represents a class of time series, to improve the computational efficiency significantly. Experimental results conducted on 26 datasets from UEA demonstrate that this algorithm achieves high classification accuracy while maintaining excellent computational efficiency.
Keywords: Multivariate time series, Shapelet, SAX representation, Time series classification.
Download PDF
1Chengdu Computer Application Institute CAS: Chengdu Information Technology of Chinese Academy of Sciences Co. Ltd., China.
2Key Laboratory of Internet Natural Language Processing of Sichuan Provincial Education Department, Leshan Normal University, China /EADDRESS/3Sichuan Provincial Key Laboratory of Philosophy and Social Science for Language Intelligence in Special Education, Leshan Normal University, China /EADDRESS/4School of Computer Science and Technology, University of Chinese Academy of Sciences, China /EADDRESS/
ABSTRACT
Time series classification (TSC) has attracted considerable attention from the time series data mining community in the last decade. In our previous work, a Compressed Random Shapelet Forest (CRSF) has been proposed to handle this issue. The CRSF algorithm not only achieves competitive performance compared with the state-of-the-art algorithms of TSC but also provides good interpretability of classification results. However, it can only support the classification of univariate time series. This paper introduces a classification algorithm, ECRSF, which extends from the CRSF algorithm for multivariate time series classification. It employs SAX to compress the time series and the Shapelet, which is a pattern that maximally represents a class of time series, to improve the computational efficiency significantly. Experimental results conducted on 26 datasets from UEA demonstrate that this algorithm achieves high classification accuracy while maintaining excellent computational efficiency.
Keywords: Multivariate time series, Shapelet, SAX representation, Time series classification.
Download PDF