CaPulse: Detecting Anomalies by Tuning in to the Causal Rhythms of Time Series

2025-08-06

AI Summary

Time-series anomaly detection faces inherent challenges including severe label scarcity, extreme class imbalance, and complex multi-periodic patterns, hindering existing methods from uncovering the underlying generative mechanisms of anomalies. To address this, we propose CaPulse, a causality-driven detection framework. First, it constructs a structured causal model to explicitly characterize the data-generating process of anomalies. Second, it introduces a period-aware normalizing flow coupled with an interpretable masking mechanism to jointly model temporal dynamics and enable anomaly root-cause attribution. Third, it incorporates a dedicated period learner for collaborative optimization. Extensive experiments across seven real-world datasets demonstrate that CaPulse achieves AUROC improvements of 3–17% over state-of-the-art methods, while exhibiting superior robustness, strong interpretability, and generalization capability.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• 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
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Time series anomaly detection has garnered considerable attention across diverse domains. While existing methods often fail to capture the underlying mechanisms behind anomaly generation in time series data. In addition, time series anomaly detection often faces several data-related inherent challenges, i.e., label scarcity, data imbalance, and complex multi-periodicity. In this paper, we leverage causal tools and introduce a new causality-based framework, CaPulse, which tunes in to the underlying causal pulse of time series data to effectively detect anomalies. Concretely, we begin by building a structural causal model to decipher the generation processes behind anomalies. To tackle the challenges posed by the data, we propose Periodical Normalizing Flows with a novel mask mechanism and carefully designed periodical learners, creating a periodicity-aware, density-based anomaly detection approach. Extensive experiments on seven real-world datasets demonstrate that CaPulse consistently outperforms existing methods, achieving AUROC improvements of 3% to 17%, with enhanced interpretability.

Problem

• Detecting anomalies in time series using causal mechanisms
• Addressing label scarcity and data imbalance in anomaly detection
• Handling complex multi-periodicity in time series data

Innovation

• Causality-based framework for anomaly detection
• Periodical Normalizing Flows with mask mechanism
• Periodicity-aware density-based detection approach