Interpret

InterpretML is an open source Python package for machine learning interpretability. It covers two jobs: training interpretable models, which it calls glassbox models, and explaining existing models that are treated as black boxes. Its flagship glassbox model is the Explainable Boosting Machine, available through a familiar scikit learn style interface, so a model can be fitted on a pandas dataframe or a numpy array with no special preprocessing. It is built for data scientists, machine learning engineers and risk or compliance teams who need to understand why a model makes the predictions it does, whether for debugging, auditing or regulated decisions. The documentation walks through a complete example on the UCI adult income dataset, from loading data and splitting it to training a glassbox model and viewing global and local explanations. The package runs on Windows, Mac and Linux and installs with pip, conda or from source. InterpretML is a code library with published documentation rather than a hosted service, and the documentation site lists no paid plans or prices.