Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection

2025-10-18

AI Summary

Real-world multivariate time series exhibit sparse, unlabeled anomalies, and existing methods suffer from overly complex architectures, coarse detection granularity, and inflated performance estimates. To address these issues, we propose OracleAD—a novel unsupervised framework that jointly models current prediction and historical reconstruction via causal embedding. It introduces Stable Latent Structure (SLS) to encode spatial relationships underlying normal behavior and employs self-attention to capture dynamic spatiotemporal dependencies. OracleAD features a dual-scoring mechanism—combining prediction error and SLS deviation—for fine-grained anomaly detection. Crucially, it localizes root-cause variables violating temporal causality directly at the embedding layer, ensuring both high accuracy and interpretability. Evaluated on multiple real-world datasets under rigorous protocols, OracleAD consistently outperforms state-of-the-art methods. It further supports both anomaly segment identification and critical variable localization.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Time-Series/Data Streams
• Reasoning under Uncertainty: Causality

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Real-world multivariate time series anomalies are rare and often unlabeled. Additionally, prevailing methods rely on increasingly complex architectures tuned to benchmarks, detecting only fragments of anomalous segments and overstating performance. In this paper, we introduce OracleAD, a simple and interpretable unsupervised framework for multivariate time series anomaly detection. OracleAD encodes each variable's past sequence into a single causal embedding to jointly predict the present time point and reconstruct the input window, effectively modeling temporal dynamics. These embeddings then undergo a self-attention mechanism to project them into a shared latent space and capture spatial relationships. These relationships are not static, since they are modeled by a property that emerges from each variable's temporal dynamics. The projected embeddings are aligned to a Stable Latent Structure (SLS) representing normal-state relationships. Anomalies are identified using a dual scoring mechanism based on prediction error and deviation from the SLS, enabling fine-grained anomaly diagnosis at each time point and across individual variables. Since any noticeable SLS deviation originates from embeddings that violate the learned temporal causality of normal data, OracleAD directly pinpoints the root-cause variables at the embedding level. OracleAD achieves state-of-the-art results across multiple real-world datasets and evaluation protocols, while remaining interpretable through SLS.

Problem

• Detects rare unlabeled anomalies in multivariate time series data
• Models temporal dynamics and spatial relationships for interpretability
• Identifies root-cause variables through stable latent structure deviations

Innovation

• Encodes past sequences into causal embeddings for prediction
• Uses self-attention to model dynamic spatial relationships in latent space
• Aligns embeddings to stable structure and uses dual scoring for anomalies