Hey,
I’m totally new to the datascience field and watched a couple of videos to get an understanding about it.
I’ve learned so far, that I can compare different models by using the ROC. Also i can use it to get the best threshold.
If I got it right, the ROC is based on the TPR (Sensitivity) and the FPR (1-Specificity). But what should I do, when there is a cost matrix for my dataset?
The TPR and the FPR are unweighted, so that the cost matrix has no influence to them. But shouldn’t it?
What can i do, to get a weighted ROC or is this just a silly question and i missed something?
Thanks for your help.
cheers
Riskit
1 post - 1 participant