Sorry in advance if my questions are really basic for some of you :
1/ while using a pretrained model (transfer learning), I am not able to pass horizon h with predicton function, why ?
use case : how can I produce a prediction for next 36 months if the trained models was trained on 24 past months ?
2/ may I finetune a model with exogenous variables while the trained model was not trained with exogenous variables ?
09/21/2023, 4:06 PM
Hi @Guillaume GALIE! The architecture of all the models in the library (expect DeepAR) depends on the horizon because they follow the "direct" forecasting strategy. When a model is trained with a particular horizon, it can't be changed afterwards.
There are two options:
1. Train the model on the longest horizon of interest, and simply trim the extra information.
2. Implement a recursive predict method, where you use the forecast as input for the next window.