I am working with a very large dataset that would benefit from using training continuation with the xgb_model parameter in xgb.train(). The label (Y) of the dataset itself has 4 classes and is highly imbalanced, so I would like to generate per-label PR curves for it to evaluate its performance, and would thus need to treat each class as its own binary problem using a one-vs-rest classifier. After a lot of reading I haven’t found an equivalent to sklearn’s OneVsRestClassifier() in the xgboost library, but thanks to this thread I understand that XGboost uses 1-vs-rest by default when dealing with multiple classes. If that’s the case, should I even bother setting an objective or a num_class param for the model? If I do need to set them, how should I think about choosing them? And secondly, if my end goal is generating per-label PR Curves (along with per-label ROC AUC curves), should I set the eval_metric directly to ‘aucpr’? Thank you in in advance for you patience and help.
How do you build 1-vs-rest classifiers in the XGBoost Library?
The easiest way to is just to use XGBClassifier.