RandomForestRegressor with monotonicity constraints

[mayer79]

Describe the workflow you want to enable

In practice, being able to use monotonic constraints in "black boxes" is very useful and a small step towards interpretability.

While quite common in gradient boosting implementations (e.g. your https://scikit-learn.org/stable/modules/ensemble.html#monotonic-cst-gbdt, XGBoost, CatBoost, LightGBM), I am not aware of any random forest implementation offering such option. I would love to see this feature in the scikit-learn random forest.

Describe your proposed solution

Same logic as for gradient boosting.

[NicolasHug]

This is tracked and addressed in #13649

[mayer79]

Oh, sweet! Should we keep this issue open? If not, I will close it.

[ItamarMushkinPagaya]

Just want to add a +1 to this very useful feature (came across this git issue while looking for monotonic constraints in Random Forest Classifiers)

[lorentzenchr]

Solved in #13649.