ce-calibrated-predict

What problem does it solve?

Calibrated predictions and probabilities are essential when model confidence is miscalibrated; this Skill provides calibrated outputs without explanations, enabling reliable decision-making with uncertainty bounds.

Core Features & Use Cases

• Calibrated point predictions via predict and calibrated probabilities via predict_proba for classification and regression tasks.
• Optional uncertainty intervals: two-sided bounds on predictions and probabilities to reflect model confidence.
• Flexible usage and integration: supports threshold-based probabilities, Mondrian-group conditioning, and works with a fully fitted and calibrated WrapCalibratedExplainer.

Quick Start

Call the explainer to return calibrated predictions for your data using predict or predict_proba, with optional uncertainty intervals.