Venue: The Fifth Conference on System and Service Quality - QualITA, May 5, 2026, Florence, Italy. Qual-ITA (https://qualitawg.github.io/) is happening with The International Conference on Performance Engineering (ICPE26) .
Presented: 11:40-12:00, 5th May, 2026 - See Program here
This repository provides the experimental results and reproducibility material for eXplainable Anomaly Detection (XAD) using the ShapBPT Python package.
ShapBPT was recently published at the
40th AAAI Conference on Artificial Intelligence (AAAI-2026), Singapore, 20–27 January 2026.
- ShapBPT Codes: https://github.com/amparore/shap_bpt
- ShapBPT Python Package: https://pypi.org/project/shap-bpt
- ShapBPT ArXiv: https://arxiv.org/abs/2602.07047
- ShapBPT Documentation: https://shapbpt.readthedocs.io/en/latest/
- ShapBPT Interactive User Study: https://huggingface.co/spaces/rashidrao/shapbpt-user-study
This repository contains the experiments, notebooks, and precomputed results for the paper:
ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study
The repository includes:
- Experimental results for AD Case Study(E6)
- Reproducible Jupyter notebooks
- Precomputed PDF and CSV files
- Ready-to-run ShapBPT examples
- Visualization utilities for anomaly detection
To run the notebooks and reproduce the experiments, create a dedicated Python environment.
conda create -n XAD python=3.9.18
conda activate XADgit clone https://github.com/rashidrao-pk/XAD
cd XAD
pip install -r requirements.txt Option 1 — Install from PyPI (Recommended) - See PyPi Project here
pip install shap-bptOption 2 — Install from Source or follow this page
git clone https://github.com/amparore/shap_bpt
cd shap_bpt
pip install .Note: ShapBPT contains a Cython module, so compilation is required when installing from source.
Some plots use LaTeX rendering.
Ubuntu/Linux:
sudo apt-get install texlive-latex-extra texlive-fonts-recommended dvipng cm-superWindows:
Install MikTeX (or equivalent LaTeX distribution).
import shap_bpt
print(shap_bpt.__version__)Download the MVTec AD dataset:
https://www.mvtec.com/company/research/datasets/mvtec-ad
Place it inside:
notebooks/datasets/
| Exp | Dataset | Model | CSV | |
|---|---|---|---|---|
| E6 | MVTec AD | VAE-GAN | PDF/HTML_E6_hazelnut_heatmaps_IoU.pdf | csv_exp_E6_testresults_hazelnut_9_BPT_new_eval.csv |
The published results were generated and validated on the following systems:
| Device | Machine | Processor | RAM | GPU |
|---|---|---|---|---|
| Laptop | Santech XN2 | Intel Core i9 (13th Gen) | 16GB | NVIDIA RTX 4070 |
| Laptop | MacBook Pro | Apple M1 | 16GB | Integrated M1 GPU |
We gratefully acknowledge the following contributions and resources that supported this work:
This work received funding from the European Union’s Horizon research and innovation programme Chips JU under Grant Agreement No. 101139769, as part of the DistriMuSe Project (HORIZON-KDT-JU-2023-2-RIA). The Joint Undertaking receives support from the European Union and the participating member states.
Our previous work (Can I Trust my Anomaly Detection, A case study base on explainable AI) Github/anomaly_detection_trust_case_study is acknowledged here which is used as pretrained model weights and initial codes.
We acknowledge the dataset MVTec curators whose work made this project possible.
Anomaly detection · Variational Autoencoder · eXplainable AI · Software quality · Smart Industries
If you find this work interesting, please cite it using following Bibtex
@inproceedings{rashid2026shapbpt,
title={ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study},
author={Rashid, Muhammad and Amparore, Elvio and Ferrari, Enrico and Verda, Damiano},
booktitle={Companion of the 17th ACM/SPEC International Conference on Performance Engineering},
pages={178--187},
year={2026}
}Note
Contributions to improve the completeness of this list are greatly appreciated. If you come across any overlooked papers, please feel free to create pull requests, open issues or contact me via email. Your participation is crucial to making this repository even better.