logo

SCIENCE CHINA Information Sciences, Volume 64 , Issue 4 : 149101(2021) https://doi.org/10.1007/s11432-020-3021-6

SPAR: set-based piecewise aggregate representation for time series anomaly detection

More info
  • ReceivedJan 1, 2020
  • AcceptedJun 4, 2020
  • PublishedFeb 26, 2021

Abstract

There is no abstract available for this article.


Acknowledgment

This work was supported by National Key Research Program of China (Grant No. U1936203), Shandong Provincial Natural Science and Foundation (Grant No. ZR2019JQ23), CERNET Innovation Project (Grant No. NGII20190109), and Project of Qingdao Postdoctoral Applied Research.


References

[1] Zhang Q, et al. Edge computing application: Real-time anomaly detection algorithm for sensing data. Journal of Computer Research and Development, 2018, 55:524--536 doi: 10.7544/issn1000-1239.2018.20170804. Google Scholar

[2] Hu Y, Ji C, Zhang Q. A novel multi-resolution representation for time series sensor data analysis. Soft Comput, 2020, 24: 10535-10560 CrossRef Google Scholar

[3] Zhan P, Sun C, Hu Y. Feature-Based Online Representation Algorithm for Streaming Time Series Similarity Search. Int J Patt Recogn Artif Intell, 2020, 34: 2050010 CrossRef Google Scholar

[4] Hu Y, Ren P, Luo W. Multi-resolution representation with recurrent neural networks application for streaming time series in IoT. Comput Networks, 2019, 152: 114-132 CrossRef Google Scholar

[5] Dau H, et al. The ucr time series classification archive, October 2018. https://www.cs.ucr.edu/~eamonn/time_series_data_2018/. Google Scholar

[6] Keogh E, Lin J, Fu A W. Finding Unusual Medical Time-Series Subsequences: Algorithms and Applications. IEEE Trans Inform Technol Biomed, 2006, 10: 429-439 CrossRef Google Scholar

[7] Keogh E, Chakrabarti K, Pazzani M. Dimensionality Reduction for Fast Similarity Search in Large Time Series Databases. Knowledge Inf Syst, 2001, 3: 263-286 CrossRef Google Scholar

[8] Ren H, Liao X, Li Z. Anomaly detection using piecewise aggregate approximation in the amplitude domain. Appl Intell, 2018, 48: 1097-1110 CrossRef Google Scholar

  • Table 1  

    Table 1Comparison experiments on 23 real world datasets

    Data set $m$ Num SAX-ADPAA-ADAPAA-ADSPAR-AD
    ADR AACI ADR AACI ADR AACI ADR AACI
    BeetleFly 512 2 0.00 0.00 0.00 0.00 1.00 2.13 1.00 2.24
    BirdChicken 512 2 0.00 0.00 0.00 0.00 1.00 2.38 1.00 2.42
    BME 128 5 1.00 2.65 1.00 3.04 1.00 2.91 1.00 4.15
    ChlorineConcentration 166 7 0.71 2.58 0.71 3.40 0.85 4.54 1.00 6.61
    Coffee 286 1 0.00 0.00 0.00 0.00 0.00 0.00 1.00 2.27
    Computers 720 0 0.00 0.00 0.00 0.00 0.00 2.17 0.00 0.00
    CricketZ 300 2 0.00 0.00 0.00 0.00 0.50 2.62 1.00 2.67
    Crop 46 3 0.33 2.17 0.67 2.20 1.00 2.39 1.00 3.21
    DiatomSizeReduction 345 3 1.00 3.81 1.00 3.90 1.00 2.98 1.00 2.79
    DistalPhalanxTW 80 7 1.00 2.48 0.85 2.49 1.00 2.76 1.00 3.99
    DodgerLoopDay 288 1 0.00 0.00 0.00 0.00 0.00 0.00 1.00 2.04
    Earthquakes 512 1 0.00 0.00 0.00 0.00 0.00 0.00 1.00 2.47
    ECG200 96 1 0.00 0.00 1.00 2.10 1.00 2.42 1.00 5.05
    ECG5000 140 2 0.00 0.00 1.00 2.42 1.00 2.47 1.00 4.03
    Fish 463 4 0.25 2.12 0.25 2.44 0.50 2.27 0.75 2.51
    GunPoint 150 3 0.00 0.00 0.33 2.23 0.33 2.12 1.00 2.66
    ItalyPowerDemand 24 1 1.00 2.22 1.00 2.39 1.00 3.68 1.00 4.44
    Meat 448 3 0.67 2.43 0.67 2.33 0.67 2.19 1.00 2.67
    OSULeaf 427 1 0.00 0.00 0.00 0.00 0.00 0.00 1.00 2.19
    Strawberry 235 2 1.00 2.38 1.00 2.54 1.00 2.76 1.00 4.03
    SDUInflowNetwork 288 2 0.00 0.00 0.00 0.00 0.50 2.11 1.00 2.97
    SDUOutflowNetwork 288 2 0.50 2.57 1.00 2.33 1.00 2.59 1.00 3.53
    SDUTotalNetwork 288 2 0.50 2.03 0.50 2.14 0.50 2.26 1.00 2.65
    Average 51% 2.49 62% 2.57 80% 2.62 98% 3.25