Deep Learning for Time Series Anomaly Detection: A Survey

TL;DR

This survey provides a structured and comprehensive overview of state-of-the-art deep learning for time series anomaly detection and provides a taxonomy based on anomaly detection strategies and deep learning models.

Abstract

Time series anomaly detection has applications in a wide range of research fields and applications, including manufacturing and healthcare. The presence of anomalies can indicate novel or unexpected events, such as production faults, system defects, or heart fluttering, and is therefore of particular interest. The large size and complex patterns of time series have led researchers to develop specialised deep learning models for detecting anomalous patterns. This survey focuses on providing structured and comprehensive state-of-the-art time series anomaly detection models through the use of deep learning. It providing a taxonomy based on the factors that divide anomaly detection models into different categories. Aside from describing the basic anomaly detection technique for each category, the advantages and limitations are also discussed. Furthermore, this study includes examples of deep anomaly detection in time series across various application domains in recent years. It finally summarises open issues in research and challenges faced while adopting deep anomaly detection models.

Figures (12)

• Figure 1: (a) An overview of different temporal anomalies plotted from the NeurIPS-TS dataset lai2021revisiting. Global and contextual anomalies occur in a point (coloured in blue). Seasonal, trend and shapelet can occur in a subsequence (coloured in red). (b) Intermetric and temporal-intermetric anomalies in MTS. In this figure, metric 1 is power consumption, and metric 2 is CPU usage.
• Figure 2: Deep Learning architectures used in time series anomaly detection
• Figure 3: General components of deep anomaly detection models in time series
• Figure 4: An Overview of (a) Recurrent neural network (RNN), (b) Long short-term memory unit (LSTM), and (c) Gated recurrent unit (GRU). These models can predict $x_t'$ by capturing the temporal information of a window of $w$ samples prior to $x_t$ in the time series. Using the error $|x_t - x_t'|$, an anomaly score can be computed.
• Figure 5: Structure of a Convolutional Neural Network (CNN) predicting the next values of an input time series based on a previous data window. Time series dependency dictates that predictions rely solely on previously observed inputs.