https://github.com/nixtla logo
#statsforecast
Title
# statsforecast
n

Nasreddine D

07/26/2023, 5:55 PM
Hi, I have tried the intermittent models (Croston...) and it seems they forecast straight lines. I'va also seen it in your tutorial. Is it the normal behaviour? Thank you
m

Mariana Menchero

07/26/2023, 6:31 PM
Hi @Nasreddine D yes, that is the usual behavior of most of the intermittent methods.
If you need something different to that, check the hierarchical forecast with non-intermittent methods for the upper levels. This can result in more variability (this assuming that your data has a hierarchical structure)
n

Nasreddine D

07/26/2023, 6:36 PM
Thank you Mariana, I will check that.
m

Max (Nixtla)

07/26/2023, 6:38 PM
Yes, this is normal and expected. Often flat lines are better than nice looking curves. The visual appeal of a generated forecast or the possibility of such a forecast to happen in general are not good criteria to judge forecasts. This is a common theme. Check this https://link.springer.com/article/10.1007/s10618-022-00894-5
n

Nasreddine D

07/26/2023, 6:39 PM
Thank you Max, I compared it with an autoArima which had a better RMSE.
m

Max (Nixtla)

07/26/2023, 6:53 PM
Nice!!! 🙂
m

marah othman

07/31/2023, 2:10 PM
so what is the best method to evaluate performance for model is it the mse and mae ? or there is something else ? @Max (Nixtla)
2 Views