Precision

A metric for classification models that answers the following question:

When the model predicted the positive class, what percentage of the predictions were correct?

Here is the formula:

where:

true positive means the model correctly predicted the positive class.

false positive means the model mistakenly predicted the positive class.

For example, suppose a model made 200 positive predictions. Of these 200 positive predictions:

150 were true positives.

50 were false positives.

In this case:

Contrast with accuracy and recall.

1.
For example, suppose a model made 200 positive predictions. Of these 200 positive predictions:
2.
were true positives.
3.
were false positives.

Created for this library

1.
A fraud team monitors precision on its declines so that customer friction from incorrect declines stays low.
2.
A medical screening team monitors precision on positive predictions because false positives drive unnecessary follow-up tests.
3.
A retention team uses precision at top-k to size the outbound call list against operational capacity.