How to plot ROC curve ?

5 vues (au cours des 30 derniers jours)
ibrahim
ibrahim le 3 Mai 2015
I work about database include 60 person and i have a result as follow : 58 identify and 2 not identify. i.e Rec rate=96.667 I ask about : 1- how to calculate tpr and fpr?.....i calculated it as tpr=58/60=0.967 and fpr=0 (is this method right?). 2- how to plot ROC curve? Note: I have only positive set= 60 person and i haven't negative set. Thanks in advance

Réponse acceptée

Image Analyst
Image Analyst le 3 Mai 2015
You don't need negatives for the ROC curve. You just need positives, and you have them both. You call them tpr and fpr. Of course since your false positive rate is 0 and doesn't vary at all, your ROC curve would be just a horizontal line going from (0, .967) to (1, .967) This good performance won't last as you begin to encounter more challenging situations. It sounds like your test set was identical to your training set so of course the ROC curve will be a very high flat line.

Plus de réponses (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by