CartoBoost EventOutcomeModel

Use EventOutcomeModel when the target is a binary event and the output should be a calibrated probability. It is for event risk, conversion, failure, completion, or acceptance probability when calibration matters.

Python Example

from cartoboost.deep import EventOutcomeModel

model = EventOutcomeModel(calibration="temperature")
model.fit(features_train, event_train)

probability = model.predict_proba(features_holdout)
report = model.calibration_report(features_holdout, event_holdout)
model.save("event-outcome.json")

Browser WASM Example

EventOutcomeModel 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

• The target is binary.
• The probability value matters, not only the class label.
• You need calibration diagnostics such as Brier score.
• A downstream decision threshold will use the predicted probability.

Use When

Need	Better first choice
Calibrated binary event probability.	EventOutcomeModel
Multiclass labels or class probabilities.	CartoBoostClassifier
Candidate-specific response curves.	ResponseCurveModel
Conformal intervals around numeric predictions.	Probabilistic and conformal models

Validation

Report Brier score, log loss, ROC-AUC or PR-AUC when appropriate, and calibration by probability bucket. Use a time, group, or entity split when deployment will face new periods or entities.

Limitations

• Calibration can drift when the event rate changes.
• A good ranking metric does not guarantee calibrated probabilities.
• Binary outcomes do not represent competing risks or time-to-event behavior.