XTREK

This repository contains the source code of XTreK and the datasets used in the experiments presented in the paper "Tree-based Kendall’s τ Maximization for Explainable Unsupervised Anomaly Detection". This paper has been accepted at the 23rd IEEE International Conference on Data Mining (ICDM 2023).

Installation

1. Install package: scikit-learn v. 1.1.2, scipy v.1.6.2, sortedcontainers v.2.4.0, and numpy v.1.21.6
2. If you want to run the unsupervised anomaly detection algorithm: Random Histogram Forest, please refer to the official repo to install: https://github.com/anrputina/rhf
3. Run experiments: python experiments_accurate.py --dataset kdd_other, dataset kdd_other is in the folder: datasets/

Usage

    from size_tree.XTREK import *
    explainer = sizebasedregressiontree(max_depth=20, max_nodes=64)
    explainer.fit(anomalyscores, x)
    explainer_predict = explainer.predict_train()