[SOLVED] help with strange ROC interpretation

makakmakak Member Posts: 13 Contributor II
Hi community,

I have a  ROC curve in AUC performance criterion (screenshot here - http://i.imgur.com/2q9F3gs.png?1) and I am not sure how to interpret it. I mean the beginning of ROC and threshold curve, it looks like first few percent of x-axis are false positives with highest confidence (1). Isn't this strange? Or I am just interpreting it wrong. Any comment, help would be very appreciated.
Thank you.


  • Options
    makakmakak Member Posts: 13 Contributor II
    Well, the interpretation was right, model really predicted many false positives with highest ranks. I have to play with model parameters...
    Sorry for spamming...
Sign In or Register to comment.