Explainable Anomaly Detection: Counterfactual driven What-If Analysis

TL;DR

This paper introduces a proof of concept for using counterfactual explanations as what-if analysis in anomaly detection, aiming to enhance interpretability and actionable insights in predictive maintenance.

Contribution

It demonstrates how counterfactual explanations can be adapted for what-if analysis in anomaly detection, providing a foundation for future complex system applications.

Findings

Implemented on PRONOSTIA dataset with TCN model

Showed counterfactuals can serve as actionable what-if scenarios

Paved the way for more advanced explainability in anomaly detection

Abstract

There exists three main areas of study inside of the field of predictive maintenance: anomaly detection, fault diagnosis, and remaining useful life prediction. Notably, anomaly detection alerts the stakeholder that an anomaly is occurring. This raises two fundamental questions: what is causing the fault and how can we fix it? Inside of the field of explainable artificial intelligence, counterfactual explanations can give that information in the form of what changes to make to put the data point into the opposing class, in this case "healthy". The suggestions are not always actionable which may raise the interest in asking "what if we do this instead?" In this work, we provide a proof of concept for utilizing counterfactual explanations as what-if analysis. We perform this on the PRONOSTIA dataset with a temporal convolutional network as the anomaly detector. Our method presents the counterfactuals in the form of a what-if analysis for this base problem to inspire future work for more complex systems and scenarios.