Ensemble of Ensembles

ensemble_methods
ensemble

#1

Is it possible or correct to create an ensemble model where each individual model itself is an ensemble. For example say I have a random forest(ensemble), a XGBoost model (ensemble) and logistic regression model. Can these be joined using voting or weighted ensemble for my final model?


#2

Hi mayanksatnalika,

Yes, you can use any of these methods after getting the predictions from individual models. More advanced algorithms used are multilevel stacking and blending. If you you want to read about these methods in detail please refer this link.

Regards
Ankit Gupta


#3

Helped a lot, thanks


#4

Hi Ankit,
Thanks a lot for the link !
I could understand the concepts explained there, however, since the implementation of those were in Python couldn’t do it practically.
Any resources for R implementations that you happen to have ?
TIA

Regards,
Giridhar K


#5

@Giridhark,

If you understand the concept there, then can’t you write your own code for doing the same? Also there are packages in R to do so, like one mentioned in this link - https://cran.r-project.org/web/packages/caretEnsemble/vignettes/caretEnsemble-intro.html

Regards,
Aayush