Data-Driven Performance in Anomaly Detection and Clustering: Evaluation Experiments

Data-Driven Performance in Anomaly Detection and Clustering: Evaluation Experiments

Context and methodology

This repository accompanies the paper When Does the Algorithm Matter? Data-Driven Performance in Anomaly Detection and Clustering and contains the experimental data and results used to evaluate anomaly detection and clustering algorithms under controlled conditions. Experiments include synthetic datasets with OFAT and factorial analyses, as well as real-world dataset collections from ADBench and ClusBench.

When Does the Algorithm Matter? Data-Driven Performance in Anomaly Detection and Clustering, by Andjela Dejelic, Tanja Zseby, and Félix Iglesias. Currently under review

Technical details

Datasets are stored in the `datasets/` directory, and experimental results are provided as CSV files in `results_Aug2026.zip`. The artifact includes Python scripts for data generation and experiment execution, together with configuration and metadata files.

The experiments require Python 3.9.6 and the dependencies listed in `docker/requirements.txt`. A pre-built Docker image and the corresponding Docker configuration are also provided to facilitate reproducibility.

Artifact Files

• datasets.zip — dataset subsets from ADBench and ClusBench collections.

• docker.zip — Docker configuration and dependency files for the reproducible environment.

• docker_image.zip — Pre-built Docker image with all required dependencies.

• FIRST_OF_ALL.md — Initial instructions and information for using the artifact.

• results_Aug2026.zip — Complete experimental results reported in the paper.

• scripts.zip — Python scripts for data generation and experiment execution.

• third_party_licenses.zip — License and attribution information for third-party datasets and resources.

Further details

See the `README.md` (within `scripts.zip`) for detailed instructions on the dataset structure, software requirements, and how to reproduce the experiments.

Licenses

• The ADBench data collection (third party) is licensed under the BSD 2-Clause
• The ClusBench data collection (third party) is licensed under CC BY 4.0 
• The results files (contained in `results_Aug2026.zip`) are licensed under CC BY 4.0 
• The rest of material and sofware files are licensed under MIT