Interpretation ROC Curve

theopilus27

I want comparing many kernel in SVM, ( dot, radial, polynomial, neural, anova, epacheninikov, gaussian combination, and multiquadric) with ROC Curve, but i dont know how i do interpretations my curve..
this is the result ROC Curve :

https://drive.google.com/file/d/0BxaT6Cn35dedRHgyS0VQMi1PWHc/edit?usp=sharing

if for can see it, please download ...

Please help me for interpretation my ROC Curve..
i know about coordinate X is FPR and coordinate Y is TPR, but i dont know how this program create value of FPR and TPR will be dot until become this curve..
please help me

MariusHelf

Hi,

does this post by wessel help? -> http://rapid-i.com/rapidforum/index.php/topic,5698.msg20162.html#msg20162

In general you can say, the ROC curve of a perfect model goes straight up to the optimum (0, 1) and then straight horizontally to (1, 1). So the higher the curve of a model, the better it is in general.


So in your case the blue and the orange line rock the ROCs

Best regards,
Marius