Hi @Germán Abrevaya if by multivariate you mean multiple series, all statsforecast's models allow you to forecast multiple series at once (just add a unique_id for them). If you mean with exogenous variables, then only some of the models allow this, namely the AutoARIMA & the MSTL-AutoARIMA. Here's a how to guide in case you haven't seen it. https://nixtla.github.io/statsforecast/docs/how-to-guides/exogenous.html
g
Germán Abrevaya
10/31/2023, 6:48 PM
Hi Mariana, thanks for your response! With multivariate, I don’t mean a single variable that is the primary one with the rest being secondary. I mean multiple coupled variables that are all equally relevant and are all targets for the forecast. For example, a dynamical system with multiple state variables.
m
Mariana Menchero
10/31/2023, 10:22 PM
ok, I guess what you need is something like a VAR model? If so, statsforecast doesn't support it yet, but if you need it please open an issue in GitHub and we'll try to work on it soon. Here's the link: https://github.com/Nixtla/statsforecast/issues
👍 1
d
DANIEL KIM
11/01/2023, 2:57 PM
Hi Mariana, if all statsforecast models allow us to forecast multiple series at once, is it possible to parallelize this process through PySpark/Spark?