search

ellisp / forecastHybrid

Convenient functions for ensemble forecasts in R combining approaches from the {forecast} package
GNU General Public License v3.0
79 stars 23 forks source link

Add Xgboost into models #51

Open nnarukulla opened 7 years ago

nnarukulla commented 7 years ago

Thank you for developing this package. This is awesome. Can we add xgboost into this hybridforecast. Most of the winning solutions for forecasting competitions having xgboost models in them. So that this package covers all time series models plus advanced machine learning algorithms (nnetar and xgboost).

dashaub commented 7 years ago

@nnarukulla For this to work, we'd have to build useful features from the timeseries so that xgboost can train on it. Do you have some example code for the types of features that xgboost trains well on? Really this might be well suited to a separate package first and then we can import it and integrate it from there.

dashaub commented 7 years ago

Also this would break our prediction intervals (which we finally gained for all the models in our ensemble).

ellisp commented 7 years ago

I've started work on a separate package at https://github.com/ellisp/forecastxgb-r-package. Very early days yet but looks promising.