InterpretML

InterpretML is an open-source Python package developed by Microsoft Research for training interpretable machine learning models and explaining black-box systems. The project represents Microsoft's comprehensive approach to machine learning interpretability, combining inherently interpretable 'glassbox' models with post-hoc explanation techniques for black-box models into a single, unified framework. The crown jewel of InterpretML is the Explainable Boosting Machine (EBM), a glassbox model that achieves accuracy comparable to state-of-the-art black-box models like XGBoost and random forests while remaining fully interpretable. EBMs are a modern implementation of generalized additive models with pairwise interactions (GA2Ms), using gradient boosting to learn smooth, non-linear feature functions while preserving the ability to visualize and understand each feature's exact contribution to predictions. This combination of high accuracy and full interpretability makes EBMs particularly valuable in regulated industries where model decisions must be fully explainable. Beyond EBMs, InterpretML provides implementations of other interpretable model types including decision rules, linear models with automatic feature engineering, and decision trees. For black-box models that cannot be replaced with interpretable alternatives, the package includes explanation methods such as SHAP, LIME, Partial Dependence Plots, and Morris Sensitivity Analysis, allowing practitioners to choose the most appropriate explanation technique for their context. InterpretML's unified API design is a significant advantage. All models and explainers follow a consistent interface, making it easy to compare interpretable and black-box approaches side by side. The package also provides an interactive visualization dashboard that renders explanations in Jupyter notebooks or standalone web views, enabling exploration of global model behavior and individual predictions through intuitive charts and graphs. The library integrates with the broader Python ML ecosystem and is compatible with scikit-learn pipelines. It is freely available under the MIT license and is actively maintained by Microsoft Research. InterpretML is particularly well-suited for organizations that want to adopt interpretable-by-design models rather than relying solely on post-hoc explanations, an approach increasingly recommended by AI governance frameworks and regulators.