This message was deleted.

Hey. Do you have the holidays as a dataframe?

I can definitely create a dataframe with the holidays. However, I was using holidays.country_holidays with subdiv specified

If you have them as a dataframe you can treat them as exogenous features (guide)

I wanna try both ahah. For the moment, I only seek to have simple features, so will only keep is_holiday.

For a single column you can define a custom date feature like this:

def is_holiday(dates):
    return dates.map(lambda x: x in holidays)

But what if holidays are not the same from a province to another one ? I wanna try to forecast on multiple time series that represent a certain location in the country

I see. Then the exogenous features approach is the best. Note that you'll have to define the dataframe by id and time, so you may have to repeat several rows

Thanks for your tips ! I'll definitely try that