Popular Posts

Sunday, May 8, 2011

Precision and Recall

One of the most important metrics used in evaluating the performance of binary classifiers is the Precision-Recall curve.

From Wikipedia:

It is possible to interpret precision and recall not as ratios but as probabilities:
* Precision is the probability that a (randomly selected) retrieved document is relevant.
* Recall is the probability that a (randomly selected) relevant document is retrieved in a search.

This page contains some illustrations to better understand the definitions of Precision and Recall:

Also, this paper relates Precision-Recall curves to the Receiver Operator Characteristic curves:

Finally, this page is further explaining the concepts and elaborating on the F-measure metric that tries to combine Precision and Recall into one measure:

No comments: