Summary Plot
Global feature importance and the direction of feature effects.
Core SHAP Visualisations
Waterfall Plot
How feature contributions move one prediction from the baseline to the final output.
Force Plot
A compact local explanation of contributions to one prediction.
Dependence Plot
How a feature value relates to its SHAP contribution.
Interaction Plot
How pairs of features jointly influence predictions.
Decision Plot
Cumulative feature contributions for one or more predictions.
Good Practice
- Use both global and local explanations.
- Check explanation stability across folds and time periods.
- Treat SHAP values as model explanations, not causal effects.
- Document the background dataset and output scale used.