I've been dabbling with a medical dataset i've managed to get my hands on and i've built a simple feed-forward network using baseline characteristics.. I've been measuring the models performance using the area under the curve. I'm calculating the auc on both the training set & test set but for some reason i'm getting a much higher auc on the test set. train-.89 test=..94
I'm a little concerned as to why theres a higher auc on the test set? If anything i'd expect it to be a little lower than the train because of overfitting