https://github.com/nixtla logo
#neural-forecast
Title
# neural-forecast
s

Sebastian Deimen

10/03/2022, 9:56 PM
Hi team, first of all, I think nhits is a great tool to work with, so, thank you for getting this out! Unfortunately, I am having trouble here and there with the documentation. I am trying to bring in holidays as exogenous data as we have more demand on those but I don't get to shape X_df correctly. I am following this colab. I am using Y_df=final_df[['unique_id', 'ds', 'y', 'time_index']] and X_df=final_df[['unique_id', 'ds', 'is_holiday']], but it throws me an error (ERRORhyperopt.fminjob exception: mat1 and mat2 shapes cannot be multiplied (154x91 and 49x256)). Also later on I want to use holiday names, as we have less contracts, eg over Christmas than over Labor Day. Would I just onehotencode those days? And how to shape that. I would appreciate some hints or pointing me at documentation where to find how X_df has to look like or any or all of it 🙂 Thanks!
c

Cristian (Nixtla)

10/04/2022, 2:05 PM
Hi @Sebastian Deimen, thanks for your interest on our N-HiTS model. We just released a new version of our library with a much easier interface. We will incorporate exogenous variables for the N-HiTS in the following days, so I suggest you to try our new interface 🙂
s

Sebastian Deimen

10/04/2022, 5:42 PM
Hey Crisitan, thank you coming back on that! Sweet, then I'll look for the new release and the new documentation in a few days! Thanks!!
m

Max (Nixtla)

10/04/2022, 10:30 PM
The first version of the new release is out 🙂
🙂 1
s

Sebastian Deimen

11/03/2022, 7:30 PM
I started playing with the new release - do I get this right that the 'auto' models are used for like hyperopt or what is the difference between those AutoNHITS and NHITS model classes?
c

Cristian (Nixtla)

11/03/2022, 7:51 PM
The 'auto' models already perform hyperparameter using Ray Tune
👍 1
3 Views