Rule Extraction in Unsupervised Anomaly Detection for Model Explainability: Application to OneClass SVM

TL;DR

This paper explores rule extraction techniques for OneClass SVM to improve its explainability in anomaly detection, introducing new algorithms and metrics to assess rule quality on real-world datasets.

Contribution

It presents novel rule extraction algorithms and metrics for assessing explainability in unsupervised models, extending XAI methods to OneClass SVM.

Findings

Effective rule extraction methods for OneClass SVM are proposed.

Metrics for comprehensibility, representativeness, stability, and diversity are introduced.

Evaluation on real-world datasets demonstrates practical applicability.

Abstract

OneClass SVM is a popular method for unsupervised anomaly detection. As many other methods, it suffers from the black box problem: it is difficult to justify, in an intuitive and simple manner, why the decision frontier is identifying data points as anomalous or non anomalous. Such type of problem is being widely addressed for supervised models. However, it is still an uncharted area for unsupervised learning. In this paper, we evaluate several rule extraction techniques over OneClass SVM models, as well as present alternative designs for some of those algorithms. Together with that, we propose algorithms to compute metrics related with eXplainable Artificial Intelligence (XAI) regarding the "comprehensibility", "representativeness", "stability" and "diversity" of the extracted rules. We evaluate our proposals with different datasets, including real-world data coming from industry. With this, our proposal contributes to extend XAI techniques to unsupervised machine learning models.