Open trungn1 opened 3 years ago
I think looking at the uncertainty intervals (yhat_lower
, yhat_upper
) for the forecast will give you a decent sense of this; they would get wider as time progresses. When the forecast has a really large uncertainty interval it's probably not useful.
What are the best practice forecasting future, what is a reasonable time window w.r.t. to your training dataset. E.g, if i only had 2 weeks of data how far in the future could I reasonable forecast for?