7.5 Model Evaluation Basics

Once a model makes predictions, we need to measure how good they are. Accuracy is the share of predictions that are correct. But on imbalanced problems, accuracy alone can mislead, so we also use precision (of the items flagged positive, how many really were) and recall (of all true positives, how many we caught).

Example: a disease test in a clinic where only 1% of people are actually sick. A model that always says 'healthy' would be 99% accurate but useless — it catches zero sick patients. Recall would expose that failure immediately.