✓Validation & evaluation

Check generalization with train-test splits, cross-validation, confusion matrices, and precision-recall. — 7 entries

Separate data used for fitting from data used for evaluation.

Rotate validation folds to measure how performance varies.

Fit training noise so closely that performance on new data degrades.

Balance systematic error against sensitivity to the training sample.

Count each kind of correct and incorrect class prediction.

Measure purity of retrieved positives and coverage of actual positives.

Plot true-positive against false-positive rates across thresholds.