← Library · Advanced concept

Recalibration of Predictive Models

Recalibration adjusts the output probabilities of a machine learning model to ensure they accurately reflect the true likelihood of an event, even if the initial model's raw scores are poorly scaled. For instance, if a model predicts a 70% chance of rain, recalibration ensures that it actually rains approximately 70% of the times the model makes that prediction. This is critical for decision-making based on probabilities.

In plain terms

It's like tuning a musical instrument so that the notes played perfectly match their intended pitch.

Why it matters

Ensures that model probabilities are trustworthy and actionable, crucial for cost-sensitive decisions or risk assessment.

Learn one new AI thing every day.

Daily Deck sends you seven plain-English cards like this every morning. Free.

Start free