shapash

Why we included this project

Shapash turns the shap and LIME outputs behind a trained model into explanations people can actually read. You generate global and per-prediction explanations in a few lines of code, then view them as clearly labeled visualizations or an interactive webapp that non-technical stakeholders can click through on their own. It also helps with auditing: you can freeze model and data information into an HTML report and check stability, consistency, and compactness metrics to see whether the explainability method itself can be trusted. That makes it a practical pick for teams working on regression, binary, or multiclass problems with XGBoost, LightGBM, CatBoost, or sklearn ensembles, especially when they need to document a model and explain its decisions to business users.