Core ML Concepts · beginner · concept 14 of 176

🧠Evaluation Metrics & Cross-Validation

How you know a model works: accuracy, precision, recall, F1, ROC curves, and cross-validation to test on data the model never saw. Choosing the wrong metric is the classic silent failure, a 99% accurate cancer detector that never says cancer is worthless. Class imbalance is the classic trap here; the fixes, resampling and class weights, live in the Data entry.