Graph-Augmented LSTM for Forecasting Sparse Anomalies in Graph-Structured Time Series
This addresses the challenge of anomaly detection in domains like sensor networks where anomalies are rare and cross-node dependencies exist, representing an incremental advance over existing methods.
The paper tackled the problem of detecting sparse anomalies in multivariate time series with relational dependencies by proposing a graph-augmented LSTM forecasting model, which improved F1-score by up to 10% over baselines on benchmark datasets.
Detecting anomalies in time series data is a critical task across many domains. The challenge intensifies when anomalies are sparse and the data are multivariate with relational dependencies across sensors or nodes. Traditional univariate anomaly detectors struggle to capture such cross-node dependencies, particularly in sparse anomaly settings. To address this, we propose a graph-augmented time series forecasting approach that explicitly integrates the graph of relationships among time series into an LSTM forecasting model. This enables the model to detect rare anomalies that might otherwise go unnoticed in purely univariate approaches. We evaluate the approach on two benchmark datasets - the Yahoo Webscope S5 anomaly dataset and the METR-LA traffic sensor network - and compare the performance of the Graph-Augmented LSTM against LSTM-only, ARIMA, and Prophet baselines. Results demonstrate that the graph-augmented model achieves significantly higher precision and recall, improving F1-score by up to 10% over the best baseline