Lesson 17 — Confusion counts

Beyond accuracy: the four outcome types

For a yes/no classifier every prediction is one of four kinds. TP (true positive): said yes, was yes. FP (false positive): said yes, was no — a false alarm. FN (false negative): said no, was yes — a miss. TN (true negative): said no, was no. From them come precision = TP / (TP + FP), "how many alarms were real", and recall = TP / (TP + FN), "how many real cases we caught".

tp = sum(1 for p, y in zip(preds, labels) if p == 1 and y == 1)

Why it matters: for a medical test, a miss (FN) is far worse than a false alarm — these counts make that visible where accuracy hides it.