I'm using Tensorflow for an auto-tagging task on audio clips. The problem is actually a multilabel classification problem meaning that each clip can have multiple tags at the same time.

Regarding the model evaluation, I would like to be able to compute the AUC per clip and per tag. My problem is that AUC is basically a metric for binary classification and I don't know how it can be extended to a multilabel problem.

I have found many references of AUC computation on multilabel classification tasks, but none of these explain how to do it.

(An example)

  • $\begingroup$ How have found any solution for this yet? $\endgroup$ May 30, 2018 at 11:37

1 Answer 1


AUC (Area under the ROC Curve) can not be calculated for multilabel classification directly.

Multilabel classification metrics have to be converted to a binary problem. The two most common conversions are:

  • OvO - One vs One
  • OvR - One vs Rest

sklearn.metrics.roc_auc_score discusses this in detail.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.