This is the official repository of the "ExIFFI and EIF+: Interpretability and Enhanced Generalizability to Extend the Extended Isolation Forest" paper.
The paper introduces Extended Isolation Forest Feature Importance (ExIFFI), a novel interpretation algorithm designed for the Extended Isolation Forest (EIF) anomaly detection model. ExIFFI aims to provide explanations for predictions made by EIF by computing global and local feature importance scores. Additionally, an enhanced variant of EIF, named EIF+, is proposed to improve generalization performance. The evaluation involves comprehensive experiments on synthetic and real-world datasets to assess anomaly detection performance and the effectiveness of ExIFFI for interpretation.
The tutorial.ipynb (contained in the notebooks folder) notebook provides a step-by-step guide on how to use ExIFFI and EIF+ for anomaly detection and interpretation. The notebook includes code snippets for training the models, making predictions, and computing feature importance scores.
Check out the Code Documentation