OptionalclassPredicted class label, present for classification problems.
OptionalcontributionsPer-record feature contributions for THIS row (P1-5) — the top signed drivers behind this specific
prediction, ranked by magnitude. Present only for models where an honest per-row attribution is cheap
and exact — linear models (coef_ · transformed value, i.e. the log-odds/value contribution). Absent
(undefined) for tree/ensemble models (which need SHAP) and multiclass; callers fall back to the model's
GLOBAL feature importance. Feature names are the post-preprocessing output columns (one-hot names like
Col=Value), so a UI should collapse/humanize them for display.
Numeric model output: probability (classification) or value (regression).
A single prediction for one input row.
scoreis the model output (probability for classification, predicted value for regression);classis the predicted label for classification problems.