Series2Graph: Graph-based Subsequence Anomaly Detection for Time Series

Abstract

Subsequence anomaly detection in long sequences is an important problem with applications in a wide range of domains. However, the approaches that have been proposed so far in the literature have severe limitations: they either require prior domain knowledge that is used to design the anomaly discovery algorithms, or become cumbersome and expensive to use in situations with recurrent anomalies of the same type. In this work, we address these problems, and propose an unsupervised method suitable for domain agnostic subsequence anomaly detection. Our method, Series2Graph, is based on a graph representation of a novel low-dimensionality embedding of subsequences. Series2Graph needs neither labeled instances (like supervised techniques), nor anomaly-free data (like zero-positive learning techniques), and identifies anomalies of varying lengths. The experimental results, on the largest set of synthetic and real datasets used to date, demonstrate that the proposed approach correctly identifies single and recurrent anomalies without any prior knowledge of their characteristics, outperforming by a large margin several competing approaches in accuracy, while being up to orders of magnitude faster. This paper has appeared in VLDB 2020.

Making Time Series Embeddings More Interpretable in Deep Learning

“…Also, different types of graph representations can be supplied as an input embedding. E. g., Boniol and Palpanas (2022) proposed a method which maps shape-related information into a graph, enabling the detection of single and recurrent anomalies. Another encoding emerged from Chengyang and Qiang (2022), where their proposed approach separates the input into frequency-based segments and builds a semantic encoded graph between those using a GCN.…”

“…One major difference in DL modeling concerns the representation of the data, where often an embedding is introduced to better cope with internal information, e. g., word2vec (Mikolov et al 2013). Recently, a lot of new time series embedding methods emerged, e. g., Kim, Hong, and Cha (2020); Chengyang and Qiang (2022); Cheng et al (2020); Yue et al (2022); Ye and Ma (2022); Tabassum, Menon, and Jastrzebska (2022); Boniol and Palpanas (2022), demonstrating that they can decrease model run time, structure information more informative, as well as improve the model performance. However, compared to words, for example, time series data is often regarded as rather complex for human interpretation.…”