Using a single model for modeling multiple time series #1194
-
|
Say, I have multiple time series of sales of different products in a store, is it possible to model them using a single NeuralProphet model? Something like generating embeddings for each product and then passing them as lagged regressors to the model? If yes, then how do I fit the model on those multiple time series? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Yes, we have something called global modeling. It is basically a single model forecasting a dataset that is comprised of multiple time series by sharing weights. |
Beta Was this translation helpful? Give feedback.
-
|
Many thanks Kevin! |
Beta Was this translation helpful? Give feedback.
Yes, we have something called global modeling. It is basically a single model forecasting a dataset that is comprised of multiple time series by sharing weights.
You can see tutorials on how to use it here and here.