Hey
@Manuel,
How are you evaluating the predictions? Or is this a qualitative evaluation?
On the side of controlling the seasonality using a multiplicative approach.
It should be possible to create a Box-Cox log scaler in the
TemporalNorm module.
Using a logarithmic transformation (multiplicative deleveling) could reduce the predictions variance.
Another idea that has given me good results is to add a
SeasonalNaive
feature to better control the seasonality in the
futr_exog_list
parameter.
Let me know your thoughts.