English

Extended Isolation Forest with feature sensitivities

Methodology 2026-02-11 v1 Machine Learning

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.

Cite

@article{arxiv.2602.09704,
  title  = {Extended Isolation Forest with feature sensitivities},
  author = {Illia Donhauzer},
  journal= {arXiv preprint arXiv:2602.09704},
  year   = {2026}
}

Comments

The automated classifier suggested cs.LG. We believe the paper is primarily machine learning theory, and we would appreciate cross-listing to cs.LG or stat.ML if deemed appropriate

R2 v1 2026-07-01T10:29:36.294Z