摘要
Time series anomaly detection through unsupervised methods has been an active research area in recent years due to its enormous potential for networks management. The representation and reconstruction of time series have made extraordinary progress in existing works. However, time series is known to be complex in terms of their temporal dependency and stochasticity, which makes anomaly detection difficult. To this end, we propose a novel approach based on a decomposition auto-transformer networks(DATN) for time series anomaly detection. The time series is decomposed into seasonal and trend components, and renovated as a basic inner block deep model. With this design, transformers can decompose complex time series in a progressive manner. We also design an auto-transfomer block that determines dependencies and representation aggregation at the sub-series level based on series seasonal and trend components. Moreover, the complex transformer decoder is replaced by a simple linear decoder, which makes the model more efficient. Extensive experiments on various public benchmarks demonstrate that our method has achieved state-of-the-art performance.
学科与关键词
- Concept Drift Adaptation for Time Series Anomaly Detection via TransformerDing, Chaoyue; Zhao, Jing*; Sun, ShiliangNEURAL PROCESSING LETTERS, 2023, 55(3): 2081-2101.被引 5
- Efficient time series anomaly detection by multiresolution self-supervised discriminative networkHuang, Desen; Shen, Lifeng; Yu, Zhongzhong; Zheng, Zhenjing; Huang, Min*; Ma, Qianli*NEUROCOMPUTING, 2022, 491: 261-272.被引 7
- Time Series Anomaly Detection with Multiresolution Ensemble DecodingShen, Lifeng*; Yu, Zhongzhong; Ma, Qianli; Kwok, James T.35th AAAI Conference on Artificial Intelligence / 33rd Conference on Innovative Applications of Artificial Intelligence / 11th Symposium on Educational Advances in Artificial Intelligence, ELECTR NETWORK, 2021-02-02 To 2021-02-09.被引 21
