InterpretML

InterpretML is an open-source Python library for data scientists who need to explain machine learning predictions or build models whose decisions they can inspect directly. It runs locally on Windows, macOS and Linux under the MIT license. Its tools support model debugging and checks for potential bias, including work in healthcare and finance.

The Explainable Boosting Machine (EBM) is its distinctive model. It combines an additive model with gradient boosting and automatic detection of interactions between features. Its explanations describe the model's actual calculations, and domain experts can edit the model. InterpretML also supports decision trees, decision rule lists and linear or logistic regression through a scikit-learn interface.

For existing models, SHAP Kernel Explainer and LIME help explain predictions without replacing the underlying model. These methods can examine entire pipelines, including preprocessing, ensembles and deep neural networks. Partial dependence and Morris sensitivity analysis provide other ways to study how inputs affect results.

Explanations cover overall model behavior and individual predictions. Notebook visualizations show feature importance and feature effects, and let users compare explanations across models. The library accepts pandas dataframes and NumPy arrays. Differentially Private EBMs are available for privacy-sensitive training; distributed EBM training for larger workloads is available separately through Azure SynapseML in the cloud.