Performing Simple Exponential Smoothing
We wish to create a simple exponential smoothing (SES) forecast with the following monthly sales values:
[33, 44, 23]
We will do this manually in order to illustrate the technique.
The level updating equation is:
Fₜ₊₁= Lₜ= αYₜ+(1-α)Lₜ-₁ = Fₜ + αeₜ
Where eₜ is the error.
Alpha is the smoothing constant and it has this property:
0 <= α <= 1
We will use an α equal to
Set Lₜ-₁ to the first observation in the data set or
33 . We could also choose another value, like the average of all the observations, or
(33 + 44 + 23)/3.
So to forecast we do:
Values Forecast (Lₜ) or (Fₜ₊₁) Error (eₜ)
44 33 +11.0
23 33.0 + 0.1 * 11 = 34.10 -11.1
34.1 + 0.1 * -11.1 = 32.99
Here we arrive at a future forecast of 32.99.