Precision and recall

Precision and recall are two fundamental measures in Machine learning and Information retrieval that describe how well a system identifies what matters.

Precision answers: of all the things your system flagged as positive, how many actually were? A spam filter with high precision rarely marks legitimate emails as junk.

Recall answers: of all the things that actually should have been positive, how many did your system catch? High recall means you're finding most of the real spam, even if you flag some false positives.

These metrics often trade off against each other. A neural network diagnosing disease might achieve high precision (few false alarms) but miss sick patients, or achieve high recall (catch all patients) but alarm healthy ones unnecessarily.

The choice between them depends on context. Medical prevention usually demands high recall—missing a disease is costly. Email filtering tolerates lower recall—some spam sneaks through, but that's acceptable.

Together, precision and recall are combined into the F1 score, a probabilistic way to balance both concerns. These concepts ripple across Machine learning, quality control, and anywhere humans need systems to separate signal from noise.

Related

Confusion matrix, Machine learning, Information retrieval, F1 score, Statistical precision, Classification (machine learning)