explainability

## What I do - Explain model predictions - Implement interpretable models - Create feature importance analysis - Build visualization tools - Design trust-building explanations - Handle regulatory requirements ## When to use me Use me when: - Model debugging - Stakeholder communication - Regulatory compliance - Bias detection - Trust in AI decisions ## Key Concepts ### Explanation Methods ```python import shap import lime from sklearn.inspection import permutation_importance # SHAP values explainer = shap.TreeExplainer(model) shap_values = explainer.shap_values(X_test) # Visualize shap.summary_plot(shap_values, X_test, feature_names=feature_names) # LIME explanations from lime.lime_tabular import LimeTabularExplainer explainer = LimeTabularExplainer(X_train, feature_names=feature_names) explanation = explainer.explain_instance(X_test[0], model.predict_proba) # Permutation importance result = permutation_importance(model, X_test, y_test, n_repeats=10) ``` ### Interpretable Models - Linear models: Coefficients - Decision trees: Path visualization - Rule-based: Explicit rules - Attention: Attention weights ### Types of Explanations - **Global**: Overall model behavior - **Local**: Single prediction - **Feature**: Feature importance - **Counterfactual**: What if scenarios