Evaluation

Model evaluation is how you measure whether a model is actually good at its job, which is trickier than a single accuracy number. If only 1% of transactions are fraud, a model that always says “not fraud” is 99% accurate and useless. So you look at precision (of the cases flagged, how many were right) and recall (of the real cases, how many were caught). Most classifiers output a score, and the threshold you pick to call something positive trades one for the other; the ROC curve shows that tradeoff across all thresholds. Calibration asks a separate question: when the model says 80%, is it right about 80% of the time.