explain ROC graph
Dhiii12
New Altair Community Member
Help me,
can anyone explain about this ROC graph on rapidminer, I'm just learning? how to get the AUC 7.511 value on this ROC chart
Thank you for your time and help!
dhiya
Tagged:
0
Best Answer
-
Hi!
ROC (Receiver Operator Characteristics) charts are described in the help of the Compare ROCs operator.
https://docs.rapidminer.com/latest/studio/operators/validation/visual/compare_rocs.html
For drawing this chart, the predictions are sorted descending by confidence. At each confidence level, the red line goes up for every correct prediction and to the right for every wrong prediction. The ideal "curve" would always go up because every decision would be right with a 100 % confidence.
The Area Under the Curve is calculated by literally determining the area under the red curve. It can be 1 as the maximum (the area of a square going from (0, 0) to (1, 1) and theoretically 0 as the minimum. 0.5 is like tossing coins: Random true or false results.
Regards,
Balázs1
Answers
-
Hi!
ROC (Receiver Operator Characteristics) charts are described in the help of the Compare ROCs operator.
https://docs.rapidminer.com/latest/studio/operators/validation/visual/compare_rocs.html
For drawing this chart, the predictions are sorted descending by confidence. At each confidence level, the red line goes up for every correct prediction and to the right for every wrong prediction. The ideal "curve" would always go up because every decision would be right with a 100 % confidence.
The Area Under the Curve is calculated by literally determining the area under the red curve. It can be 1 as the maximum (the area of a square going from (0, 0) to (1, 1) and theoretically 0 as the minimum. 0.5 is like tossing coins: Random true or false results.
Regards,
Balázs1 -
0
-
two question please,what is the difference between blue line and red line,0
-
Hi!
As you can see in the charts, the blue line is called "ROC (Thresholds)". These are the confidence levels as described.
The areas are the variance resulting from cross validation iterations.
Regards,
Balázs1