For problems with binary classification, roc auc curve or roc auc score is often used to rate a model. But does the ROC ACC make sense in the context of a binary classification model that outputs only 0 or 1 for each observation (so no probability between [0,1] but only 0, 1)
$\begingroup$
$\endgroup$
2
-
$\begingroup$ What is ROC ACC CURE and ROC ACC? Or is it AUC? $\endgroup$– gunesCommented Nov 10, 2019 at 18:27
-
$\begingroup$ @gunes sry, roc auc score $\endgroup$– Felix HaCommented Nov 10, 2019 at 18:35
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
If the classifier, at its final step, doesn't produce a probability, and only makes $0,1$ predictions; ROC doesn't make sense because in order to draw a meaningful curve you need to sweep a threshold value. If all the values are either $0$ or $1$, sweeping the threshold makes no changes doesn't change the assigned classes, and has the same TPR, FPR for all the intermediary thresholds.