AUC seems too high, confusion matrix seems only slightly better than random

Question

My confusion matrix looks as follows:

> table(actual, predicted_all)
      predicted_all
actual    0    1
     0 1728 5261
     1 2088  168

While the AUC seems a bit too high in my eyes:

auc(actual, predicted_all) Area under the curve: 0.8391

I suspect I'm doing something wrong here, or is this not the case?

Answer 1

You have 6989 samples in class 0, 2256 samples in class 1, so it is easy to beat a random predictor, i.e. the AUROC should be significantly higher than 0.5.

This confusion matrix was simply computed using a threshold that makes your predictor tends to choose 1 more often.