An evaluation metric for classification. Precision (positive predictive value) is the fraction of predicted positives that are actually correct. In trading classification (e.g. buy vs. sell signals), precision answers: “Of all signals our model labeled as buy, what proportion led to a real up movement?” High precision means few false buy signals. Formally, Precision = TP/(TP+FP). Precision is important when false signals are costly; it is often paired with recall to fully assess a classifier.

Back to Robots Glossary Index