In this page, I will introduce how to set holidays data for time series forecasting.
We provide a data extension to get bank holidays, so let's import data from it.
It can be used from this "Import Extension Data" dropdown menu.
If you haven't installed the extension yet, please add it from "Add New" button. Then, the plugin is added to "Installed" and you can import data from "Import Data".
You can choose holiday type, start year and end year.
If you import the data, it will be like this.
do_prophet, which is the function for time series forecast, expect the name of the date column to be "ds", so let's rename it.
And it also need "holiday" column, so let's create it.
If you'd like to to have the effect of holiday before or after the holiday too, you can also add "lower_window" column and "upper_window" column. For example, if you want the holiday effect to take effect from 1 day before the holiday to 1 day after the holiday, you can set -1 to the "lower_window" column, and 1 to the "upper_window" column. For more detail on the content of the data frame for holidays, refer to the official document of Prophet.
Then, you can choose this data frame from Time Series Forecast dialog and the dates in the data frame are recognized as holidays.