SHAP (SHapley Additive exPlanations)
VerifiedOpen-source Python library for model interpretability using Shapley values. Explains individual predictions and global model behavior. 22,000+ GitHub stars.
Frameworks Covered
NIST-AI-RMF
Services
model-explainabilityfeature-importanceinterpretabilitybias-detection
About SHAP
SHAP (SHapley Additive exPlanations) is a game-theoretic approach to explain the output of any machine learning model. It connects optimal credit allocation with local explanations using the classic Shapley values from cooperative game theory.
Features
- Model-agnostic and model-specific explainers
- Local and global interpretability
- Feature importance visualization
- Interaction effects between features
- Supports tree-based models, deep learning, and linear models
- Waterfall, force, and summary plots
- Bias detection via feature attribution analysis
Installation
pip install shap
Loading reviews...