https://github.com/nixtla logo
l

lbfan

07/06/2023, 5:15 AM
I'm wandering when I train a time series prediction model which will have a history input length of
n
, and prediction length of
m
For a training set, for each time series group with different unique_id but have a much longer length than
n
, will the
model.fit()
method automatically slide window training by the training set?
k

Kevin Kho

07/06/2023, 5:58 AM
Which library are you using? Statsforecast?
c

Cristian (Nixtla)

07/06/2023, 4:10 PM
Hi @lbfan! Yes, during training windows of size
n
are sampled at random in each training iteration.
l

lbfan

07/07/2023, 1:31 AM
That's great, thanks @Cristian (Nixtla)