AutoARIMA fit takes a long time

[matsuobasho]

I'm just starting with statsforecast.

I have univariate time series data with a frequency of 15minutes. I subsetted it to just 3 months of data (11K records).

sf = StatsForecast(
    models = [AutoARIMA(season_length = 96)],
    freq = '15min'
)

sf.fit(df_comb[df_comb.ds>='2022-09-01'])

It ran for 87 minutes on my machine before I killed it. Is this normal?

[jmoralez]

The ARIMA model doesn't scale very well with the season length, Rob Hyndman suggests creating fourier terms instead to model the seasonal patterns, you can find an example on how to do that for statsforecast here

[matsuobasho]

Thanks, @jmoralez for the quick help. Will try fourier terms as per your suggestion.