MAIF/shapash

Shapash is a Python library for machine learning model interpretability and explainability that generates interactive visualizations and web applications to make model behavior understandable to both technical and non-technical audiences.

The library addresses the challenge of understanding how machine learning models make decisions by building on SHAP and LIME methodologies to provide clear, labeled explanations. It generates an interactive web application that lets users explore feature interactions and navigate between local explanations for individual predictions and global explanations of overall model behavior. The tool also produces comprehensive audit reports that document model and data characteristics, helping teams understand and validate their models before deployment.

Shapash works with regression, binary classification, and multiclass problems across numerous model types including Catboost, Xgboost, LightGBM, Sklearn ensemble methods, linear models, and SVM. It is designed for data science teams who need to share model insights with stakeholders ranging from fellow data scientists to business users without machine learning expertise. The web application interface is the primary distinguishing feature, enabling interactive exploration rather than static visualizations alone.

The project maintains active test coverage through continuous integration workflows. Development activity shows consistent engagement with the codebase through regular updates and maintenance of the documentation site. The team actively solicits user feedback through structured channels to inform the direction of the library.