Anomaly Detection
Finding rare rows that do not belong to normal patterns, often a fraud or failure class, scored by precision and recall rather than accuracy alone.
Think of It Like This
Like an airport luggage scanner that rings an alarm on the one bag containing an odd metallic shape among thousands of ordinary suitcases.
With a rare base rate, a detector that catches most real anomalies can score lower raw accuracy than a baseline that flags nothing, since accuracy averages over both classes and the majority class's size dominates. Precision and recall, scored on the rare class specifically, are what actually measure whether a detector is doing its job.
Statistical, proximity-based, and model-based methods all assume normal behavior is common and anomalous behavior is rare, differing only in how they characterize that norm.