How to choose a threshold in a binary classification problem?
Feb 8, 2021
In the binary classification problem, we may get the output when we have pretty good ROC AUC, however, the accuracy is pretty bad when we set the threshold to 0.5.
A nice discussion about the case when we get high AUC with low accuracy can be found here.
Can we improve the accuracy by adjusting the threshold? We may be allowed to do so.
