question: how can you use future values of weather? about auckland_cycling HOT 2 OPEN

Hi @ihadanny, sorry for the rather late reply. in reply to your questions:
1: I here use observed values for the weather variables, but in the context of e.g. a full predictive model which would aim at predicting the cyclist numbers a few days in advance, you simply would use the outputs of a numerical weather prediction model
2: because of the nature of the model, you cannot really do cross validation, as you need to respect the chronological ordering of the time-series. Not sure about your comment related to an overfit, the model does not 'see' the test data at any point, and the train vs test validation metrics are quite similar

from auckland_cycling.

1. got it
2. https://robjhyndman.com/hyndsight/tscv/

from auckland_cycling.