Scikit-Explain Documentation

scikit-explain is a user-friendly Python module for machine learning explainability. For a comprehensive tutorial, see Flora et al. (2024).

Current explainability products include:

• Feature Importance:

  • Single- and Multi-pass Permutation Importance (Breiman et al. 2001; Lakshmanan et al. 2015; McGovern et al. 2019)

  • First-order PD/ALE Variance (Greenwell et al. 2018)

  • Grouped Permutation Importance (Au et al. 2021)

• Feature Effects/Attributions:

  • Partial Dependence (PD)

  • Accumulated Local Effects (ALE)

  • Individual Conditional Expectations (ICE)

  • SHapley Additive Explanations (SHAP)

  • Local Interpretable Model-Agnostic Explanations (LIME)

  • Random forest-based feature contributions (TreeInterpreter)

• Feature Interactions:

  • Second-order PD/ALE

  • Interaction Strength (IAS) and Main Effect Complexity (MEC) (Molnar et al. 2019)

  • Friedman H-statistic (Friedman and Popescu 2008)

  • Sobol Indices

A primary feature of scikit-explain is the accompanying plotting methods, which are designed to be easy to use while producing publication-quality figures. Computations leverage parallelization when possible.

The package is under active development. Feel free to raise issues!

Citation

If you employ scikit-explain in your research, please cite:

@article{Flora_2024,
  author  = {Flora, Montgomery L. and McGovern, Amy and Handler, Shawn},
  title   = {A Machine Learning Explainability Tutorial for Atmospheric Sciences},
  journal = {Artificial Intelligence for the Earth Systems},
  volume  = {3},
  number  = {1},
  pages   = {e230018},
  year    = {2024},
  doi     = {10.1175/AIES-D-23-0018.1},
  url     = {https://journals.ametsoc.org/view/journals/aies/3/1/AIES-D-23-0018.1.xml}
}

Installation

pip (PyPI):

pip install scikit-explain

conda (conda-forge):

conda install -c conda-forge scikit-explain

Development version (most up-to-date):

git clone https://github.com/monte-flora/scikit-explain.git
cd scikit-explain
pip install -e .

Tutorials

• Quickstart
  • Compute Permutation Importance
  • Compute ALE Curves
  • Next Steps
• ExplainToolkit API
  • ExplainToolkit
  • PlotConfig
• Feature Importance
  • Permutation Importance
  • Grouped Importance & Comparing Methods
• Feature Effects
  • Accumulated Local Effects (ALE)
  • Partial Dependence (PD)
  • Individual Conditional Expectations (ICE)
  • 2D Feature Effects
• Feature Attributions
  • Local Attributions (SHAP, LIME, TreeInterpreter)
  • SHAP-Style Plots
• Feature Interactions
  • Feature Interactions
• Multiclass Classification
  • Multiclass Classification
• Plot Configuration
  • Plot Configuration
  • API Reference

Contribute

• Issue Tracker: https://github.com/monte-flora/scikit-explain/issues

• Source Code: https://github.com/monte-flora/scikit-explain

Support

If you are having issues, please let us know. We have a mailing list located at: monte.flora@noaa.gov

License

The project is licensed under the BSD license.