Three ways of being right, and the reason ‘ninety-nine per cent accurate’ can be worthless. If one transaction in a thousand is fraudulent, a system that says ‘not fraud’ every single time is 99.9 per cent accurate and completely useless. Precision asks: of the cases it flagged, how many were real? Recall asks: of the real cases, how many did it catch? You can nearly always trade one against the other, and which matters is a business decision rather than a technical one.
Technical view
Precision is true positives over predicted positives; recall is true positives over actual positives. The trade-off is controlled by the decision threshold applied to predicted probabilities and is best examined through a precision-recall curve, which is more informative than a receiver operating characteristic curve under heavy class imbalance. Accuracy is only meaningful when classes are reasonably balanced and the two error types cost about the same. When screening a large population, even a very low false positive rate produces a large absolute number of wrong flags, which is usually the binding constraint.
Related
Part of AI Terminology Index, section: Measuring Performance.