Solved

Getting the seasonality for FORECAST_ETS function

  • 14 September 2023
  • 2 replies
  • 365 views

Userlevel 2
Badge +1
  • Seasoned Pigmenteer
  • 22 replies

Dears,

I have the historic sales per month, and we have strong seasonality, but also and upward trend. I would like to use the FORECAST_ETS function, however, I don’t understand how to get the value of seasonality. For example I know every July and August we have a drop and then it gets higher on September. In this case the seasonality would be 12?

 

If I want something more exact can I use Python and then get the equation and plug in pigment? 

 

For calculating a better seasonality number, alpha, beta and gamma, is there a way to get those numbers in Pigment, or I need to have another application, such as Python or R?

 

Thanks in advance!

 

Jose

icon

Best answer by Jurgen Maas 31 January 2024, 11:41

View original

2 replies

Userlevel 6
Badge +10

hey Jose,

Forecast_ETS is modeled after the one of excel which is a simple exponential smoothing.

For very seasonal series, I recommend Triple Exponential Smoothing or Holt-Winters.
We don’t have functions for these but both are doable with Pigment metrics and math operators.

Userlevel 2
Badge +3

You can create it in Python and the output send to Pigment via the API.

Or what I did was to build a custom statistical forecast application in Pigment.
There I used the logic of “Multiplicative Decomposition” (https://en.wikipedia.org/wiki/Decomposition_of_time_series) to identify the real seasonality and trend.

 

Reply