Advanced Windowing Techniques

Hello TensorPeople!

I am working on a time-series model and I based the program off of the tensorflow time-series tutorial, so I have the same WindowGenerator class that uses the same ‘split_window’ function to map a tensoflow dataset. I’m wondering if anyone know if its possible to pass in forecasted features into the model.

i.e. - I have forecasted data, and I need to be able to pass in the forecasted features next to the labels that need to be predicted.

Its also a multi-step and a multi-output model, so there are 4 labels.

For some reason, this forum wouldnt let me post pictures, so I have some diagrams of what I’m trying to explain here, in my google drive.

normal_model

model_w_forecasted_feature_as_input

Are you looking for multi-step?

Hi Bhack, thank you for your reply. I’m looking for something a little more specific. I have included some visuals below:

normal_model

model_w_forecasted_feature_as_input