Extended Isolation Forest with feature sensitivities

TL;DR

This paper introduces the Anisotropic Isolation Forest, an extension of the standard isolation-based anomaly detection method that allows for adjustable feature sensitivities and directional sensitivity measures, improving detection focus.

Contribution

It presents the first adaptation of isolation forests to incorporate feature sensitivities and directional measures, enabling more targeted anomaly detection.

Findings

AIF effectively detects anomalies in preferred feature directions.

AIF outperforms standard EIF on synthetic and real datasets.

Directional sensitivity measures facilitate task-specific tuning.

Abstract

Compared to theoretical frameworks that assume equal sensitivity to deviations in all features of data, the theory of anomaly detection allowing for variable sensitivity across features is less developed. To the best of our knowledge, this issue has not yet been addressed in the context of isolation-based methods, and this paper represents the first attempt to do so. This paper introduces an Extended Isolation Forest with feature sensitivities, which we refer to as the Anisotropic Isolation Forest (AIF). In contrast to the standard EIF, the AIF enables anomaly detection with controllable sensitivity to deviations in different features or directions in the feature space. The paper also introduces novel measures of directional sensitivity, which allow quantification of AIF's sensitivity in different directions in the feature space. These measures enable adjustment of the AIF's sensitivity to task-specific requirements. We demonstrate the performance of the algorithm by applying it to synthetic and real-world datasets. The results show that the AIF enables anomaly detection that focuses on directions in the feature space where deviations from typical behavior are more important.