Open coforfe opened 3 years ago
Thanks for the heads up. I could probably write some pseudo-code here, but lightgbm
can be added by the end user just like any other model from any other package.
Based on past experience, training lightgbm
models in parallel over multiple cores may not work when it comes time to predict. Perhaps I was serializing the model incorrectly, but this is just a heads up.
Thanks Nick for the info.
I was not aware of this restriction for lighgbm
when predicting.
Perhaps it is an issue to clarify with the authors of the package. I will ask them and update it here when I have their response.
Thanks again! Carlos.
Hi Nick,
lightgbm
is finally available in CRAN as another type of boosting algorithm.Since it can compete in performance and speed with
xgboost
. It could be useful to include it as another forecaster algorithm you can choose from.Thanks, Carlos.