In pattern recognition, information retrieval, and binary classification, precision, also known as positive predictive value, is the ratio of true positives, or correctly identified class members, to the total number of positive results or total identified class members. Consider the following table:

Negative Positive
Actual Negative TN FP
Positive FN TP
TN (True Negative)  ≡  # of actual negative results identified correctly as negative
FN (False Negative) ≡  # of actual positive results mis-identified as negative (Type 2 errors)
FP (False Positive)  ≡  # of actual negative results identified mis-identified as positive (Type 1 errors)
TP (True Positive)  ≡  # of actual positive results identified correctly as positive

The precision is the number of true positive results divided by the number of all positive results:

precision = \dfrac{TP}{TP + FP}


« Back to Glossary Index

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.