Plotting explanations

ShapBPT provides a built-in visualization helper:

shap_bpt.plot_owen_values(
    explainer,
    shap_values,
    class_names,
    names=None,
)

Interpretation

The plotting utility overlays attribution maps for the explained classes and uses the original image for context.

Typical workflow

shap_values = explainer.explain_instance(1000, method="BPT")
shap_bpt.plot_owen_values(explainer, shap_values, class_names)

Tips

• Use consistent class names with the model outputs.

• Compare BPT and AA results side by side when evaluating the effect of the hierarchy.

• Save figures explicitly with Matplotlib when preparing reports or papers.