CartoBoost ResponseCurveModel

Use ResponseCurveModel when rows contain context features, candidate values, and observed responses, and you want to estimate how response changes as the candidate value changes. This is useful for price, threshold, bid, dose, offer, or service-level curves.

Python Example

from cartoboost.deep import ResponseCurveFrame, ResponseCurveModel

frame = ResponseCurveFrame.from_pandas(
    candidates,
    feature_cols=["region_feature", "time_feature", "entity_feature"],
    candidate_value_col="candidate_value",
    response_col="response",
    group_col="decision_id",
    candidate_id_col="candidate_id",
)

model = ResponseCurveModel(
    response_type="binary",
    monotone="decreasing",
    calibration="isotonic",
    backend="cpu",
)
model.fit(frame)

curve = model.predict_curve(frame)
response = model.predict_response(frame)
best = model.best_candidate(frame)

Browser WASM Example

ResponseCurveModel Wasm example

Runs the Rust-backed browser export on a substantial multi-route taxi panel with a held-out prediction window.

Ready to run in this page.

When To Use

• Candidate value is a controlled input, not just another feature.
• You need a curve or best candidate, not only one prediction per row.
• The candidate effect should be monotone increasing or decreasing.
• Candidate groups represent decisions that should be compared together.

Use When

Need	Better first choice
Candidate response curves.	ResponseCurveModel
Calibrated binary probability without a candidate curve.	EventOutcomeModel
Select one feasible candidate from scored rows.	ConstrainedDecisionOptimizer
Generic row-level regression.	CartoBoostRegressor

Validation

Use grouped validation so candidates from the same decision do not leak across train and holdout. Report response metrics and whether the selected candidate improves the decision metric against simple candidate rules.

Limitations

• Observational candidate-response curves are not automatically causal.
• Monotonic constraints encode assumptions that must be justified before fitting.
• Extrapolation beyond observed candidate values is unreliable.