ROCAUC and Cross Validation

[bjschoenfeld]

We have decided to just use sklearn's cross validate functionality and set return_estimator=True. With the estimator, we can call predict twice, once to get the scores we now compute and once to get the probabilities to compute roc auc metrics.