Describe the workflow you want to enable

Hi,

While ensemble methods like RandomForestRegressor are widely used, scikit-learn currently lacks native support for estimating and exposing predictive uncertainty—an increasingly essential feature in many applied domains such as healthcare, scientific modeling, and decision support systems.

Describe your proposed solution

I propose adding functionality to expose both:

Aleatoric uncertainty (data-driven),
Epistemic uncertainty (model-driven).

Importantly, this is not just a concept—I have already implemented this wrapper as part of my ongoing PhD research. The approach is detailed in a preprint available here:

http://dx.doi.org/10.22541/au.175373261.14525669/v1 .

The implementation is functional, tested, and used in geophysical mapping described in the paper.

This contribution builds on established research by Mohammad Hossein Shaker and Eyke Hüllermeier in uncertainty estimation for Random Forest Classification, and I have extended those principles to Random Forest Regression.

The approach is detailed in this article available here:

http://dx.doi.org/10.1007/978-3-030-44584-3_35

Thanks

Describe alternatives you've considered, if relevant

Additional context

No response