In various graph neural network (GNNs) papers, the ROC-AUC metric is usually shown alone without considering F1 or Accuracy.

  1. Is there a reason for that?
  2. What does it say about two models 1 and 2 with scores auc_1 and auc_2, where auc_1 > auc_2? (besides saying that maybe model 1 has better performance across various thresholds than model 2)
  3. How does one go about choosing the right model to get the best performance on accuracy let's say, given that data is balanced?

Thank you



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.