interpret

The library centers on Explainable Boosting Machine (EBM) a state-of-the-art glass-box model that combines the accuracy of gradient boosting with full interpretability, enabling users to understand exactly why a model makes each prediction without post-hoc approximation.

Beyond EBM, InterpretML provides implementations of classical interpretable models (linear regression, logistic regression, decision trees, rule lists) alongside post-hoc explainability methods for black-box models SHAP, LIME, Partial Dependence Plots, Morris Sensitivity Analysis, and saliency maps for neural networks.

The unified API allows practitioners to switch between model types and explanation methods with minimal code changes, and an interactive visualization dashboard renders explanations as interactive charts for model debugging and stakeholder communication.

InterpretML is open-source under the MIT license and is particularly relevant for regulated industries where model explainability is a legal or compliance requirement financial services (credit scoring, loan decisions), healthcare (diagnostic support, treatment recommendations), and hiring automation where algorithmic transparency is mandated.

The EBM model class is notable for achieving competitive accuracy with XGBoost and LightGBM on tabular data benchmarks while remaining fully interpretable, challenging the common assumption that accuracy and interpretability are fundamentally in tension.