jinlow / forust

A lightweight gradient boosted decision tree package.
https://jinlow.github.io/forust/
Apache License 2.0
56 stars 6 forks source link

Add variable importance measures #37

Closed jinlow closed 1 year ago

jinlow commented 1 year ago

Look at the different variable importance measures supported by XGBoost and add those calculations to forust.

gravesee commented 1 year ago

This paper could be useful for implementing an unbiased feature importance measure

https://arxiv.org/pdf/2305.10696.pdf

jinlow commented 1 year ago

Basic feature importance measures have now been implanted (gain, cover, and weight). Closing this, and will open a separate ticket to keep track of the unbiased gain.