I have a sales dataset with 4 columns, Date, StoreID, ItemID and Sales. In the dataset, there are 100 stores and each store having 20 items for sales.
I want to do sales forecasting for each store-item combination. But my challenge is what model should I use? Should I use the same time series model for each store-item combination or I should use different models? There are 2000 store-item combination, seems like it is not practical to have 2000 time series models for each unique combination.
If I used one model for all the 2000 combinations, then what parameters should I use for the model since different combination are having different seasonal pattern, stationary etc... Should I use additive or multiplicative for the seasonal mode?
I'm quite confuse on how to do the forecasting for my scenario, and I couldn't find any resource that I can refer to. Any help or advise will be greatly appreciated. Please share with me if you have any resource/documentation that are similar to my scenario. Thank you.