Conformal Anomaly Detection in Event Sequences

Shuai Zhang, Chuan Zhou, Yang Liu, Peng Zhang, Xixun Lin, Shirui Pan

TL;DR: We propose a novel conformal anomaly detection method for event sequences, which combines two newly designed non-conformity scores with provably valid p-values for hypothesis testing.

Abstract: Anomaly detection in continuous-time event sequences is a crucial task in safety-critical applications. While existing methods primarily focus on developing a superior test statistic, they fail to provide guarantees regarding the false positive rate (FPR), which undermines their reliability in practical deployments. In this paper, we propose CADES (Conformal Anomaly Detection in Event Sequences), a novel test procedure based on conformal inference for the studied task with finite-sample FPR control. Specifically, by using the time-rescaling theorem, we design two powerful non-conformity scores tailored to event sequences, which exhibit complementary sensitivities to different abnormal patterns. CADES combines these scores with Bonferroni correction to leverage their respective strengths and addresses non-identifiability issues of existing methods. Theoretically, we prove the validity of CADES and further provide strong guarantees on calibration-conditional FPR control. Experimental results on synthetic and real-world datasets, covering various types of anomalies, demonstrate that CADES outperforms state-of-the-art methods while maintaining FPR control.

Lay Summary: Anomaly detection in event sequences is essential for ensuring safety in areas like healthcare, finance, and information security, where unexpected sequences can have serious consequences. Existing methods typically focus on detecting these anomalies, but they do not offer rigorous control over false positives, meaning they might mistakenly flag normal sequences as anomalies. In this paper, we introduce a new method called CADES, based on a statistical framework called conformal prediction, which improves anomaly detection by controlling false positives. Additionally, CADES can identify anomalies that current methods fail to detect. Through a series of experiments, we show that CADES outperforms existing methods while maintaining control over false positive rate. This approach provides a more trustworthy way of detecting anomalies, which is crucial for applications where safety is a top priority.

Link To Code: https://github.com/Zh-Shuai/CADES

Primary Area: General Machine Learning->Sequential, Network, and Time Series Modeling

Keywords: Conformal Inference, Anomaly Detection, Event Sequence, Temporal Point Process

Submission Number: 3842