Prediction interval with XGBoost about autoquant HOT 4 CLOSED

@Shafi2016 I'm not currently doing anything with the AutoH2OModeler() function. I have quantile regression available with that function if you use h2o gbm or h2o DL. You can also use AutoH2oGBMRegression() and AutoCatBoostRegression() for quantile regression by setting the Alpha argumenty, although I believe h2o returns better results. I do not have any current plans to add in quantile regression for the xgboost versions.

from autoquant.

Thanks, Adrian for the prompt response. Let me check it with the AutoH2oGBMRegression(). If I can get what I am looking for. I have the same conclusion that h2o XGBoost provides better accuracy and is much better than the other packages. I checked almost every package both in python and R.

from autoquant.

@Shafi2016 I know with XGBoost I just need to set up a pinball loss function and then XGBoost can deliver quantile regression. I just haven't spent the time on that at this point.

from autoquant.

That will be a good addition. As currently, this is not available for XGBoost in any package of R and Python.

from autoquant.