You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I understand from the README that it supports multiple covariates. And I understand I can make multiple models, one for each time series, from another GitHub issue here. Does it support multiple time series datasets, also known as multi-instance?
As a more concrete example, consider this OJ Sales Dataset. In this case, the data contains weekly sales of orange juice over 121 weeks. There are 3,991 stores included and three brands of orange juice per store so that 11,973 models can be trained.
I understand one can train independent models for each of the stores. However, I was interested in knowing if the package can take in data from multiple stores to learn correlations between them.
The text was updated successfully, but these errors were encountered:
If you are referring to 'global modelling' aka shared model weights across all stores.
The feature is not well documented yet, but should become clearer end of week. (See answer in #531)
No
If you are looking to do 'VAR' - explicitly correlating all stores and all products (likely not what you plan to do).
Yes
If you are looking for a way to regress one store's product sales onto the same or another stores product sales.
Here's what I would do:
Create 3 dataset collections, with one product as the target series each.
Within each collection, create one dataset for each store.
Add the store's other two items as lagged regressors (additional columns in dataframes)
Add the 1-10 closest stores' same item sales as additional lagged regressors (additional columns in dataframes)
Train a global model for each of your 3 products. (I'd use local normalization once Local normalization for global modeling #507 is merged).
Note: the input will need to be a dictionary containing each dataset as an entry.
Now you can use the 3 models to predict all products at all stores.
Note: the input will need to be a dictionary containing each dataset as an entry.
I understand from the README that it supports multiple covariates. And I understand I can make multiple models, one for each time series, from another GitHub issue here. Does it support multiple time series datasets, also known as multi-instance?
As a more concrete example, consider this OJ Sales Dataset. In this case, the data contains weekly sales of orange juice over 121 weeks. There are 3,991 stores included and three brands of orange juice per store so that 11,973 models can be trained.
I understand one can train independent models for each of the stores. However, I was interested in knowing if the package can take in data from multiple stores to learn correlations between them.
The text was updated successfully, but these errors were encountered: