In statistical classification, the recall of a classifier over a class is the ratio between the number of times a class got predicted correctly and the number of times that class actually appears in the corresponding dataset.

recall(x)=Ncorrectly predicted(x)N(x), \text{recall}(x) = \frac{N_{\text{correctly predicted}}(x)}{N(x)},

where xx is the class being classified.

Typically, this appears in binary classification, where we want to compare the number of true positives against the number of positives.

See also Precision, Accuracy and F-score.