Do deep neural networks contribute to multivariate time series anomaly detection?

作者：

Highlights：

• We study the performance of three family of methods, conventional, machine learning and deep learning ones, for anomaly detection in multivariate time series

• No family of methods shows significant outperformance over the five datasets considered for evaluation

• Deep learning methods seem to perform better on time series containing contextual anomalies

• Conventional techniques outperform the other methods when the available training data is small

摘要

•We study the performance of three family of methods, conventional, machine learning and deep learning ones, for anomaly detection in multivariate time series•No family of methods shows significant outperformance over the five datasets considered for evaluation•Deep learning methods seem to perform better on time series containing contextual anomalies•Conventional techniques outperform the other methods when the available training data is small

论文关键词：Anomaly detection,Multivariate time series,Neural networks

论文评审过程：Received 6 September 2021, Revised 31 May 2022, Accepted 25 July 2022, Available online 26 July 2022, Version of Record 30 July 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108945