Is there a reason why Rulefit does not support XGBoost ?

[pangjacque]

Hi,

I understand that sklearn's tree structure is straightforward to traverse: each GradientBoostingClassifier stores its trees in .estimators_, and each tree has a .tree_ attribute, while xgboost tree structure is incompatible with the sklearn .tree_ interface.

However, we can always use get_dump() or similar get_booster() API to extract and adapt for XGBoost tree format.

Other than engineering simplification reason, is there a design decision not embracing XGboost ? Are we open for a PR to embrace XGBoost into Rulefit ?

Thanks a lot!

[csinva]

Very open to it, just haven't had time to get around to it!