interpret

Why we included this project

InterpretML is a Python package built around a simple premise: you should be able to see why a model makes a prediction before you put it in front of users. Its flagship model, the Explainable Boosting Machine (EBM), is an additive tree ensemble developed at Microsoft Research that matches gradient-boosted trees on accuracy while producing exact explanations for both the model's overall behavior and individual predictions. The same library also includes blackbox explainers such as SHAP and LIME, so models trained elsewhere can be audited too. That coverage matters in regulated settings like healthcare or finance, where debugging mistakes and checking for bias count as much as raw accuracy, and where compliance reviews expect documentation you can actually read. It works with the familiar scikit-learn API, accepts string features without extra encoding, and offers a differentially private version of EBM when the data is sensitive.