X-Why

XWhy: eXplain Why with SMILE -- Statistical Model-agnostic Interpretability with Local Explanations

Installation

pip install xwhy

Simple Example

import xwhy
import xgboost

# train an XGBoost model
X, y = xwhy.datasets.boston()
model = xgboost.XGBRegressor().fit(X, y)

# explain the model's predictions using xwhy
# (same syntax works for LightGBM, CatBoost, scikit-learn, transformers, Spark, etc.)
explainer = xwhy.Explainer(model)
xwhy_values = explainer(X)

# visualize the first prediction's explanation
xwhy.plots.waterfall(xwhy_values[0])

Citations

It would be appreciated a citation to our paper as follows if you use X-Why for your research:

@article{Aslansefat2021Xwhy,
   author  = {{Aslansefat}, Koorosh and {Hashemian}, Mojgan and {Martin}, Walker and {Papadopoulos}, Yiannis},
   title   = "{SMILE: Statistical Model-agnostic Interpretability with Local Explanations}",
   journal = {arXiv e-prints},
   year    = {2021},
   url     = {https://arxiv.org/abs/...},
   eprint  = {},
}

Acknowledgment

This project is supported by the Secure and Safe Multi-Robot Systems (SESAME) H2020 Project under Grant Agreement 101017258.

Contribution

If you are interested in contributing to this project, please check the contribution guidelines.