Closed mbanco closed 3 years ago
@mbanco Selecting the confidence levels is a great idea. I think it should be straightforward to add.
What do you mean by adding the forecast function though? Returning the model?
Thanks Adrian, This is the forecast function included in the forecast package:
forecast( object, h = ifelse(frequency(object) > 1, 2 * frequency(object), 10), level = c(80, 95), xreg = NULL, ... )
This allows modifying the confidence level or the value of the regressors, etc without having to refit the model.
@mbanco Can I just return the model object and the xreg object as well? If you have those two then you can run the forecast function without refitting
I can also have them saved to file which might be the easiest thing to do. That way, you can just load them into R when you need them.
If you can return the model object and the xreg object, that would be very good!!
@mbanco Let me start by adding a Path argument to the functions so that the model and xregs can be saved to a file location of your choice if a path is supplied. I can do that quickly.
@mbanco You can now supply a path to the FilePath arg and the model will be saved and if there are XREG and XREGFC objects, those too (or two) will be saved to file.
Let me know if that works for both requests. With the model output you can reforecast and supply different prediction interval values.
Thank you very much, Adrian! I'm going to try this version.
Regards
El mié, 10 feb 2021 a las 20:57, Adrian (notifications@github.com) escribió:
@mbanco https://github.com/mbanco You can now supply a path to the FilePath arg and the model will be saved and if there are XREG and XREGFC objects, those too (or two) will be saved to file.
Let me know if that works for both requests. With the model output you can reforecast and supply different prediction interval values.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/AdrianAntico/RemixAutoML/issues/71#issuecomment-777120837, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2ZWYVASHXC3W6SVIEFZG3S6MMPTANCNFSM4XLDAAEQ .
Let me know how it goes. I think you will need to run RemixAutoML::TimeSeriesDataPrepare() to get the training data to run the forecast() function. I think the PerformanceGrid output from the AutoBandit() and Auto() functions will let you know how to set the params in TimeSeriesPrepare(). Let me know if you can't get that to work..
Hi Adrian, Would it be possible to allow the AutoArfima, AutoTBATS, AutoBanditNNet and AutoBanditSarima functions to include the possibility of modifying the confidence level and also include the forecast function, as in the forecast package?
Best regards,
Mauricio