With Time Series Forecast as a step, you can specify a holiday table. In holiday table, you can specify types of holidays, and effect of each type of holiday is estimated separately.

Here is data on access count of the Wikipedia page “Apple World Wide Developers Conference”.

Let’s make forecast on this data, taking effects from 2 different types of holidays into account.

  1. Regular holidays on calendar.
  2. The days WWDC (Apple World Wide Developers Conference) is held. This is not really a holiday, but a type of days that would affect the result.

Here is the data on holidays.

The column name has to be ds and holiday. holiday column is the type of holiday. “NYSE holiday” is regular holidays on calendar, and “wwdc” is the WWDC days.

Then, make a forecast as a step, with the holiday data frame specified as Data Frame for Holidays.

In the output, you will see effect of NYSE holiday,

and, effect of WWDC in separate columns.

You can visualize them in a line chart, with separate lines for effect of each type of holidays.

Here is an interactive chart.