There was a paper written on a new AUC metric for multiclass problems.
http://proceedings.mlr.press/v97/kleiman19a/kleiman19a.pdf
The authors have provided the code in the repository below so this should be straightforward to incorporate as an inbuilt metric.
For the reference: catboost has already implemented AUC Mu: https://github.com/catboost/catboost/commit/298ca12c7c7c8bda54ef1291b3fb1ea0888e6539.
Closed in favor of being in #2302. We decided to keep all feature requests in one place.
Welcome to contribute this feature! Please re-open this issue (or post a comment if you are not a topic starter) if you are actively working on implementing this feature.
PR is open for this.
Closed via #2567.
Most helpful comment
PR is open for this.