Performance metric for multi-class classification

Question

Hi all

I've just trained a multi-class classification model with random forest.

I have split train-test sets.

Apart from Area Under Curve (AUC), what other performance metrics can I use to diagnose/improve the model (?)

There are 24 levels for my target variable, I don't think confusion matrix is a good measure (?)

Thanks

Answer 1

Have you tried this :
https://sebastianraschka.com/faq/docs/multiclass-metric.html

This is the first result when you google your subject.