Learning Graph Neural Networks for Multivariate Time Series Anomaly Detection

TL;DR

This paper introduces GLUE, a graph neural network model that improves multivariate time series anomaly detection by learning variable dependencies and quantifying predictive uncertainty, leading to more accurate and interpretable results.

Contribution

GLUE extends GDN by incorporating uncertainty estimation and automatic dependency learning, enhancing anomaly detection and interpretability in multivariate time series.

Findings

GLUE's forecasting performance matches or exceeds GDN and vector autoregressor baselines.

GLUE provides meaningful sensor embeddings that cluster similar sensors.

Uncertainty estimation improves anomaly detection robustness.

Abstract

In this work, we propose GLUE (Graph Deviation Network with Local Uncertainty Estimation), building on the recently proposed Graph Deviation Network (GDN). GLUE not only automatically learns complex dependencies between variables and uses them to better identify anomalous behavior, but also quantifies its predictive uncertainty, allowing us to account for the variation in the data as well to have more interpretable anomaly detection thresholds. Results on two real world datasets tell us that optimizing the negative Gaussian log likelihood is reasonable because GLUE's forecasting results are at par with GDN and in fact better than the vector autoregressor baseline, which is significant given that GDN directly optimizes the MSE loss. In summary, our experiments demonstrate that GLUE is competitive with GDN at anomaly detection, with the added benefit of uncertainty estimations. We also show that GLUE learns meaningful sensor embeddings which clusters similar sensors together.