Can neural prophet model available for multivariates?

[landkwon94]

First of all, many thanks for the contributions!

I have a question about this model usage.

I am studying all tutorials in this github, and I see that most tutorials are about uni-variate forecasting.

As I want to perform multivariate time series forecasting, I guess these codes :

col_lst = ['precipitation', 'soil_moisture', 'solar_radiation', 'temperature', 'wind_speed']
m = m.add_lagged_regressor(names=col_lst, normalize="minmax")

may train the model for multivariate forecasting.

May I ask am I correct?
And is there any tutorials which deal with multi-variate? I want to study if there is :)

Another question is that, is it possible to add regional information for training the model?
For example, not only numeric values (e.g., precipitation, temperature, etc.), I want to add regions (e.g., State, Country, etc.) for training.

I guess tutorial in here : https://github.com/ourownstory/neural_prophet/blob/main/docs/source/tutorials/tutorial09.ipynb
may be related to my question. But just curious, is it possible to train the model with regional values?

Again, super thanks for your prompt reply!!!

[leoniewgnr]

Hi @landkwon94, thanks for your question.
Multivariate forecasting
Yes, NP can do multivariate forecasting. This tutorial describes how to add regressors: https://neuralprophet.com/tutorials/tutorial05.html
You can just do it like the code your code snippet.

Regional information
That is also possible. We call it a global local model. Because some parts are modelled globally (over all regions) and some local (individual for each region). Feel free to check out this tutorial: https://neuralprophet.com/tutorials/tutorial09.html

Hope this helps! Let me know if you have further questions!