But roc = round(roc_auc_score(y_test, test_scores))
is for supervised learning,
In the above example, if there is no ground truth, can we use roc auc? if yes,how?