When using `ARIMA`/`AutoARIMA` in statsforecast, you can model a single seasonality (SARIMA) or multiple seasonalities using

MSTL

. You can also try

AutoARIMAProphet

which relies on Fourier terms like FB Prophet to model multiple seasonalities (importing the relevant functions from Prophet itself). However,

AutoARIMAProphet

uses a Prophet-like API and not a nixtla/statsforecast-like API. It would be nice if there was a way to use Fourier-based seasonalities with a statsforecast-like API. Maybe something like pmdarima's FourierFeaturizer https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.preprocessing.FourierFeaturizer.html could be implemented in statsforecast or in the Nixtla/tsfeatures package? They could also be useful as exog variables for some neuralforecast models.

Yeah, we can look into that. Could you open an issue for this on Github?