F1 is a classification metric that combines precision and recall into a single score using their harmonic mean. It is useful when both false positives and false negatives matter, and it tends to penalize large imbalances between precision and recall.
A "roll-up" binary classification metric that relies on both precision and recall. Here is the formula: Suppose precision and recall have the following values: - precision = 0.6 - recall = 0.4 You calculate F~1~ as follows: When precision and recall are fairly similar (as in the preceding example), F~1~ is close to their mean. When precision and recall differ significantly, F~1~ is closer to the lower value. For example: - precision = 0.9 - recall = 0.1