Dealing with Uncertainty in Contextual Anomaly Detection

TL;DR

This paper introduces a novel uncertainty-aware framework for contextual anomaly detection that models both aleatoric and epistemic uncertainties, improving detection accuracy and interpretability, especially in critical domains like healthcare.

Contribution

The paper proposes a new framework called normalcy score (NS) for CAD that explicitly models uncertainties using heteroscedastic Gaussian process regression, enhancing detection reliability.

Findings

NS outperforms existing CAD methods in accuracy

Confidence intervals improve interpretability and decision-making

Effective in real-world healthcare applications

Abstract

Contextual anomaly detection (CAD) aims to identify anomalies in a target (behavioral) variable conditioned on a set of contextual variables that influence the normalcy of the target variable but are not themselves indicators of anomaly. In many anomaly detection tasks, there exist contextual variables that influence the normalcy of the target variable but are not themselves indicators of anomaly. In this work, we propose a novel framework for CAD, normalcy score (NS), that explicitly models both the aleatoric and epistemic uncertainties. Built on heteroscedastic Gaussian process regression, our method regards the Z-score as a random variable, providing confidence intervals that reflect the reliability of the anomaly assessment. Through experiments on benchmark datasets and a real-world application in cardiology, we demonstrate that NS outperforms state-of-the-art CAD methods in both detection accuracy and interpretability. Moreover, confidence intervals enable an adaptive, uncertainty-driven decision-making process, which may be very important in domains such as healthcare.