(Copied from old kiwi) |
m |
||
| (One intermediate revision by one other user not shown) | |||
| Line 8: | Line 8: | ||
In applications like information retrieval, one can usually trade off recall and precision by using another measure called 'F-measure'. The expressions for these terms can be found at [http://cs.haifa.ac.il/~shuly/teaching/04/statnlp/supervised.pdf this site] | In applications like information retrieval, one can usually trade off recall and precision by using another measure called 'F-measure'. The expressions for these terms can be found at [http://cs.haifa.ac.il/~shuly/teaching/04/statnlp/supervised.pdf this site] | ||
| + | |||
| + | [[Category:ECE662]] | ||
Latest revision as of 07:46, 10 April 2008
Recall and Precision Metrics
Estimation of classifiability of various Pattern Recognition techniques is needed to compare them against each other and also to select the most suited one for a given application. The metrics called 'Recall' and 'Precision' are borrowed from Information Retrieval for this estimation.
'Recall' is the proportion of target items that the system selected. It is used to specify the completeness of the classifier where as 'Precision' is the proportion of the selected items that the system got right. It is used to specify the correctness and accuracy of the classifier.
The idea will be clear by following the usage of these metrics in information retrieval, for example in a web search engine. (Wikipedia:Performance measures)
In applications like information retrieval, one can usually trade off recall and precision by using another measure called 'F-measure'. The expressions for these terms can be found at this site
