Why is changing only boostingtype to rf insufficient to configure LightGBM Random Forest mode?
Classic MLLevel: IntermediateRandom forest mode configuration
Why is changing only boosting_type to rf insufficient to configure LightGBM Random Forest mode?
Hints
- How are training samples formed and how do ensemble trees differ?
- Consider correlation between model errors, not only individual quality.
Answer
The FAQ specifies sampling settings as part of Random Forest mode: use bagging_fraction and feature_fraction below one and enable bagging through bagging_freq. These control row and feature subsampling; an arbitrary boosting configuration is not automatically valid for rf. Check the constraints for the installed version and validate the result on held-out data. This is a configuration issue, not evidence that Random Forest and gradient boosting are the same algorithm.
Source and licence: LightGBM developers — Official FAQ · MIT · answer: Quantoros editorial