Classical models built from the math you already know
A smoke detector that never goes off is right almost every single day, because most days there is no fire. Call that detector "99% accurate" and the number is technically true, and completely useless, because the one day it matters, a real fire, is exactly the day it stays silent. Classifiers trained on real data fall into the same trap constantly, because most real classification problems are not a fair fifty-fifty split. Fraud is rare. A specific disease is rare. A defective part on a factory line is rare. Almost every example belongs to the ordinary class, and the one worth catching is the minority.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ Imbalanced Classes: When One Label Dominates