Welcome to Francis Academic Press

Academic Journal of Computing & Information Science, 2024, 7(4); doi: 10.25236/AJCIS.2024.070407.

Research on Improving Time Series Similarity Based on Segmented Local Representation

Author(s)

Jing Zhou, Mingwei Li

Corresponding Author:
Jing Zhou
Affiliation(s)

Faculty of Science, Northeastern University, Shenyang, 110819, China

Abstract

The similarity measurement of time series has important application value in multiple fields. Among them, the approximate representation and similarity measurement of sequences are the key to solving similarity search. The main purpose of this study is to solve the problem that traditional similarity measures cannot capture the similarity between long time series well. By using segmented local representation and the loss function based on local representation of time series, the loss distance is proposed as an indicator for similarity measurement. The method proposed in this paper is to match the traditional point-to-point according to the time scale to the local representation matching between the current segments according to the morphological characteristics of the time series, which saves the time cost and improves the efficiency of the model.

Keywords

Segmented local representation; time series similarity; Loss distance

Cite This Paper

Jing Zhou, Mingwei Li. Research on Improving Time Series Similarity Based on Segmented Local Representation. Academic Journal of Computing & Information Science (2024), Vol. 7, Issue 4: 45-55. https://doi.org/10.25236/AJCIS.2024.070407.

References

[1] Liu Boning, Zhang Jianye, Zhang Peng, et al. Time series similarity search method based on curvature distance [J]. Journal of Electronics and Information Technology, 2012, 34 (9): 2200-2207

[2] Li Hailin, Guo Chonghui. Symbolic Aggregation Approximation Method for Time Series Based on Morphological Features [J]. Pattern Recognition and Artificial Intelligence, 2011,24 (5): 665-672

[3] Liu Fen, Guo Gongde. A composites measurement method for time series similarity based on symbolic aggregation approximation [J]. Computer Application, 2013,33 (1): 192-198

[4] Xiao Rui, Liu Guohua. Trend based similarity measurement and clustering research for time series [J]. Computer Application Research, 2014,31 (9): 2600-2605

[5] Keogh E, Lonardi S, Ratanamahatana C A. Towards parameter-free date mining[C].Proceedings of the 10th ACM SIGKDD international conference on Konwledge discovery and date mining. ACM, 2004: 206-215

[6] Keogh E, Ratanamahatana C A. Exact indexing of dynamic time warping [J]. Knowledge and Information Systems, 2005, 7(3): 358-386. 

[7] Gorecki T, Luczak M. Using derivatives in time series classification [J]. Data Mining and Knowledge Discovery, 2013, 26(26): 310-331. 

[8] Keogh E J, Pazzani M J. Derivative dynamic time warping[C]//Proceedings of the 1st SIAM International Conference on Data Mining, Chicago, Apr 5-7, 2001. Philadelphia: SIAM, 2001: 1-11

[9] Nakamura T,  Taki K,  Nomiya  H,  et  al.  A  Shape-based  Similarity  Measure  for  Time  Series  Data with  Ensemble  Learning[J].  Pattern Analysis and Applications, 2013, 16(4): 535-548. 

[10] Wang Da, Rong Gang. Pattern Distance of Time Series [J]. Journal of Zhejiang University (Engineering Edition), 2004,38 (7): 795-798

[11] Xie Fuding, Li Ying, Sun Yan, et al. Improved Symbolic Time Series Processing Method [J]. Computer Engineering and Design, 2012,33 (10): 3950-3953