SHAP Library Reference¶

SHAP connects game theory with machine learning to provide robust, globally consistent explanations.

Quick API¶

TreeExplainer: Optimised specifically for Tree-based models (XGBoost, Random Forest). Extremely fast.
```
explainer = shap.TreeExplainer(model)
```
LinearExplainer: For linear models (Logistic Regression, Linear Regression).
```
explainer = shap.LinearExplainer(model, X_train)
```
KernelExplainer: The fallback for any model (including neural networks or complex ensembles). Very slow.
```
explainer = shap.KernelExplainer(model.predict, X_train_summary)
```

shap.plots.waterfall(shap_values[0]): Explains a single prediction.
shap.plots.force(shap_values[0]): Alternative horizontal view for a single prediction.
shap.plots.beeswarm(shap_values): Global view showing feature intensity and impact across the whole dataset.
shap.plots.bar(shap_values): Standard global feature importance.

KSB	Description	How This Addresses It
S5	Deployment, value assessment, and ROI	Translating model performance into business impact
S6	Communicate through storytelling and visualisation	Presenting ML results to non-technical stakeholders
B4	Consideration of organisational goals	Framing technical results in terms of business objectives
B1	Inquisitive approach	Exploring creative ways to explain model behaviour