自編碼器與動態閾值用於單變量時間序列之異常偵測

Anomaly Detection in Univariate Time Series with An Autoencoder and Dynamic Thresholding

張廷維(Ting-Wei Zhang)

指導教授 ： 王勝德

國立臺灣大學/電機資訊學院/電機工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002924

摘要

物聯網中的傳感器設備常以時間序列的形式提供數據，例如橋樑振動，溫度，人體生理數據和空氣品質。本文提出了一種可以同時考慮時間序列的重建特徵和時間依賴性之異常檢測模型。所提出的模型基於帶有預測網絡的自動編碼器，該網絡可以即時計算每個時間戳上的異常分數。此外，考慮到每個時間序列之間的異常分數可能會根據環境因素而變化，我們設計了一個動態閾值演算法來為每個單變量時間序列提供一個個別的動態閾值。我們所提出帶有動態閾值演算法的深度學習模型在YahooWebscope數據集中的A1真實基準和Numenta異常基準（NAB）數據集上取得了良好的結果。

關鍵字

異常偵測 ； 時間序列 ； 動態閾值 ； 自編碼器 ； 非監督

並列摘要

Sensor devices in Internet of Things often provide data in the form of time series, such as bridge vibrations, temperatures, human physiological data and air quality. The thesis proposes an anomaly detection model that can simultaneously considers the reconstruction feature and temporal dependence oftime series. The proposed model is based on an auto encoder with a prediction network, which can instantly calculate the anomaly score at each time stamp. In addition, to consider that the anomaly scores among each time series may vary according to environmental factors, we designed a dynamic threshold algorithm to provide an individual dynamic threshold for each univariate time series. The proposed deep learning model with the dynamic threshold algorithm has been shown to achieve good results on the A1 real benchmark in the Yahoo Webscope dataset and Numenta Anomaly Benchmark (NAB) dataset.

並列關鍵字

Anomaly Detection ； Time Series ； Dynamic Threshold ； Autoencoder ； Unsupervised

參考文獻

[1]Chuxu Zhang, Dongjin Song, Yuncong Chen, Xinyang Feng, Cristian Lumezanu, Wei Cheng, Jingchao Ni, Bo Zong, Haifeng Chen, and Nitesh Chawla. A deep neural network for unsupervised anomaly detection and diagnosis in multivariate time series data. Proceedings of the AAAI Conference on Artificial Intelligence, 33:1409–1416, 07 2019.

[2]M. Munir, S. A. Siddiqui, A. Dengel, and S. Ahmed. Deepant: A deep learning approach for unsupervised anomaly detection in time series. IEEE Access, 7:1991– 2005, 2019.

[3]Mohammad Braei and Sebastian Wagner. Anomaly detection in univariate timeseries: A survey on the state-of-the-art, 2020.

[4]V. Hautamaki, I. Karkkainen, and P. Franti. Outlier detection using k-nearest neighbour graph. In Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004., volume 3, pages 430–433 Vol.3, 2004.

[5]Markus Breunig, Hans-Peter Kriegel, Raymond Ng, and Joerg Sander. Lof: Identifying density-based local outliers. volume 29, pages 93–104, 06 2000.