Precision in Machine Learning

Definition

With precision, we get a sense of: when we classified an item as class j, was it actually class j?

Precision is high when we are careful about assigning class j to an item. For example, imagine there are 100 samples that are class j. Even if we only label 10 of these as class j and do not incorrectly label any other items as class j, then our Precision = 1 which is maximal (here TP = 10 and FP = 0). Mentally, we can replace the term TP with Correctly-Labeled-as-Class-j and FP with Incorrectly-Labeled-as-Class-j.

TP = True Positive FP = False Positive

\[ \text{Precision} = \frac{TP}{TP + FP} \]

References

Recall in Machine Learning