California Consumer Privacy Act (CCPA)
Model Evaluation
The alignment between a model's predicted probabilities and the actual observed frequencies of events.
Detailed Explanation
Calibration measures how accurately a model's predicted probabilities reflect real-world outcomes. A well-calibrated model's predicted chances, such as a 70% probability, correspond closely to the actual occurrence rate, like 70 out of 100 times. Proper calibration ensures reliable decision-making, especially in sensitive applications like medical diagnosis or risk assessment.
Use Cases
- •Use case: Ensuring medical risk scores accurately reflect patient outcomes to improve diagnosis reliability and treatment decisions.
Related Terms
Other terms in the Model Evaluation category