scikit-explain

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

Explainability Methods

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)
• TreeInterpreter (tree-based feature contributions)

Feature Interactions

• Second-order PD/ALE
• Interaction Strength (IAS) and Main Effect Complexity (MEC) (Molnar et al. 2019)
• Second-order PD/ALE Variance (Greenwell et al. 2018)
• Second-order Permutation Importance (Oh et al. 2019)
• Friedman H-statistic (Friedman and Popescu 2008)
• Sobol Indices

These methods are discussed in Christoph Molnar's Interpretable Machine Learning. A primary feature of scikit-explain is the built-in plotting methods, designed to be easy to use while producing publication-quality figures. Documentation is available at Read the Docs.

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 .

Dependencies

scikit-explain is compatible with Python 3.8 or newer and requires:

numpy, scipy, pandas, scikit-learn, matplotlib, shap>=0.30.0,
xarray>=0.16.0, tqdm, statsmodels, seaborn>=0.11.0

Quick Start

import skexplain

# Load pre-trained models and data
estimators = skexplain.load_models()
X, y = skexplain.load_data()

# Create the explainer
explainer = skexplain.ExplainToolkit(estimators=estimators, X=X, y=y)

# Configure plot display settings once (optional)
explainer.set_plotting_config(
    display_feature_names={"sfc_temp": "$T_{sfc}$", "temp2m": "$T_{2m}$"},
    display_units={"sfc_temp": "$^\\circ$C", "temp2m": "$^\\circ$C"},
)

Permutation Importance

perm_results = explainer.permutation_importance(n_vars=10, evaluation_fn='norm_aupdc')
explainer.plot_importance(data=perm_results, panels=[('multipass', 'Random Forest')])

Accumulated Local Effects

important_vars = explainer.get_important_vars(perm_results, multipass=True, nvars=7)
ale = explainer.ale(features=important_vars, n_bins=20)
explainer.plot_ale(ale=ale)

Feature Attributions

import shap

single_example = X.iloc[[0]]
explainer = skexplain.ExplainToolkit(estimators=estimators, X=single_example)

shap_kws = {
    'masker': shap.maskers.Partition(X, max_samples=100, clustering="correlation"),
    'algorithm': 'auto',
}
attr_results = explainer.local_attributions(
    method=['shap', 'lime', 'tree_interpreter'],
    shap_kws=shap_kws,
)
explainer.plot_contributions(attr_results)

Tutorial Notebooks

Notebook	Description
01 Quickstart	Minimal workflow from model to explanation
02 Permutation Importance	Single/multi-pass permutation importance
03 Grouped Importance	Grouped PI and comparing ranking methods
04 ALE	1D Accumulated Local Effects
05 Partial Dependence	1D Partial Dependence
06 ICE Curves	Individual Conditional Expectations
07 2D Effects	2D ALE and Partial Dependence
08 Local Attributions	SHAP, LIME, and TreeInterpreter
09 SHAP Plots	Summary and dependence plots
10 Interactions	H-statistic, IAS, MEC, Sobol indices
11 Multiclass	Multiclass classification support
12 Plot Configuration	Customizing plots with PlotConfig

Citation

If you use 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},
}

Acknowledgments

This package includes adapted code from: PyALE, PermutationImportance, ALEPython, SHAP, scikit-learn, LIME, Faster-LIME, treeinterpreter

Contributing

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

License

BSD license.