GCAD: Anomaly Detection in Multivariate Time Series from the Perspective of Granger Causality

2025-01-23

AI Summary

To address inaccurate causal modeling and limited interpretability in multivariate time series anomaly detection, this paper proposes an end-to-end deep learning framework grounded in Granger causality. Departing from conventional prediction- or reconstruction-based paradigms, our method dynamically infers time-varying causal structures among variables by backpropagating gradients through a nonlinear deep predictor; it further constructs a lightweight causal graph via adaptive sparsification. Crucially, we are the first to fully embed Granger causal discovery into a differentiable deep learning pipeline, enabling gradient-driven evolution of the causal graph. Evaluated on multiple real-world datasets, our approach consistently outperforms state-of-the-art baselines, achieving an average F1-score improvement of over 8%. Moreover, it provides interpretable causal pathways underlying detected anomalies—thereby unifying high detection accuracy with strong causal interpretability.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Multivariate time series anomaly detection has numerous real-world applications and is being extensively studied. Modeling pairwise correlations between variables is crucial. Existing methods employ learnable graph structures and graph neural networks to explicitly model the spatial dependencies between variables. However, these methods are primarily based on prediction or reconstruction tasks, which can only learn similarity relationships between sequence embeddings and lack interpretability in how graph structures affect time series evolution. In this paper, we designed a framework that models spatial dependencies using interpretable causal relationships and detects anomalies through changes in causal patterns. Specifically, we propose a method to dynamically discover Granger causality using gradients in nonlinear deep predictors and employ a simple sparsification strategy to obtain a Granger causality graph, detecting anomalies from a causal perspective. Experiments on real-world datasets demonstrate that the proposed model achieves more accurate anomaly detection compared to baseline methods.

Problem

• Multivariate Time Series
• Anomaly Detection
• Causal Relationships

Innovation

• Granger Causality
• Multivariate Time Series
• Anomaly Detection