In most of the places, I have found that sensitivity=recall. In terms of the Confusion Matrix, the formula for both of these is the same: $TP/(TP+FN)$. Is there any difference between these two metrics? If not, then why does the same thing has a different name?
$\begingroup$
$\endgroup$
3
asked Aug 15, 2018 at 13:36
-
4$\begingroup$ The reason is probably its use in different fields. Recall is more of a production term and sensitivity more of a clinical term. The terms are also more self-explanatory in those fields, so I guess that's why. $\endgroup$– Frans RodenburgCommented Aug 15, 2018 at 13:39
-
1$\begingroup$ True positive rate is yet another name for the same thing. In my opinion it is the most logical, because you can infer the definition from it: the amount of true positives with respect to all positives. $\endgroup$– sebpCommented Aug 15, 2018 at 13:46
-
$\begingroup$ This may make it harder to communicate ideas, especially with audiences that may lack much background in statistics, but I find it useful to think of these as $F_{\hat Y\vert Y = y}$. This generalizes the idea of sensitivity/recall beyond binary classification, even to problems with continuous outcomes. Then $F_{Y\vert \hat Y = \hat y}$ generalizes precision (positive predictive value). $\endgroup$– DaveCommented Oct 30 at 0:29
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
It is not uncommon that statistical tools have different origins and names, but same meaning.
The name sensitivity comes from the statistics domain as a measure for the performance of a binary calssification, while recall is more related to the Information Engineering domain.
$\begingroup$
$\endgroup$
1
Marsland suggests that they are only the same when there are only two classes. (It's on p. 25 of Machine Learning: An Algorithmic Perspective (2E), and yes, there is a typo, but one can mentally correct that read the rest of the sentence that ends 2.2.6.)
-
4$\begingroup$ I think this is potentially useful, but not quite as yet. Can you say more about what the difference is in the multiclass case? How would the meaning differ? When would each be more useful? Could you create a simple confusion matrix that illustrates the issues? Could you quote the sentence from the book? Etc. $\endgroup$ Commented Oct 29 at 19:58