forecastHybrid
forecastHybrid copied to clipboard
ellisp
Convenient functions for ensemble forecasts in R combining approaches from the {forecast} package
The current 'forecast' method from the forecastHybrid package does not allow new ts data to be used with an existing model. In contrast the 'forecast' from the forecast package method...
Could we add the prophet model into the list of models that are fit and combined in the end? It performs very well for daily data with weekly and annual...
Hi, I don't know if there is a problem but it seems that the `cv `function struggle to handle the `xreg `argument if we choose an `hybridModel `as `FUN `...
I would be nice to add the `arfima` model from forecast package. Such "model" are quite uncommon but are working quite well for times series that have long rum memory...
Currently `cvts()` only saves residuals in the object return. The actual or fitted values will need to be included and the `accuracy.cvts()` method will need to be modified to implement...
My suggestion is to create an matrix with all possible combinations of like Then a function can create an matrix with all possible combinations e.g: ``` aut
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...
Currently `forecast.hybridModel()` issues a warning if the `xreg` was only used for one component model and not the others. This behavior could be kept, but more flexibility would be given...
Add the ability to update the `weights`/`errorMethod` on an existing hybridModel object. For `errorMethod` this can easily be achieved without recalculating the component models since the RMSE/MAE/MASE errors are all...
We should consider the option of separate weights to each forecast horizon (e.g. `auto.arima` performs very well for short horizons, `nnetar` for medium horizons, and `ets` for long horizons, so...