shap

What problem does it solve?

SHAP provides a principled, scalable way to explain model predictions by attributing output to individual features using Shapley values. It makes machine learning decisions more transparent, trustworthy, and actionable.

Core Features & Use Cases

• Global & local explanations: Compute SHAP values for any model and visualize with waterfall, beeswarm, bar, and scatter plots.
• Model debugging & fairness: Debug model behavior, check for bias, and compare models.
• Production-ready explainability: Integrate explanations into dashboards or APIs to support decision-making.

Quick Start

Train a tree-based model (e.g., XGBoost), create a TreeExplainer, compute SHAP values on your test data, and visualize a global importance plot such as a beeswarm.