Single Exponential Smoothing Forecast Method
This forecast method calculates a weighted moving average forecast by applying exponentially decreasing weights to the demand observation as they go farther in the past, thus the most recent demand observations get the highest weight.
Some things to consider:
• This forecast method uses a smoothing parameter α (alpha) [0,1] to smooth the demand slice by slice and calculate forecast.
• The higher the value of Alpha, the more weight is given to the recent demand slices, but higher values of alpha can also cause high variation in month to month forecast.
• This forecast method is more appropriate when trend and seasonality are not significant factors.
• This forecast method produces a flat-line forecast. Single Exponential Smoothing requires at least one month of demand history.
The Single Exponential Smoothing forecast method calculates the forecast as follows:
F(t+1) = α * D(t) + (1 – α) * F(t)
F(2) = D(1)
Where:
• t = [2, # History Slices]
• α is the smoothing parameter
• D(t) is the Demand history for period ‘t’
• F(t) is the Forecast
 |
Each forecast method has constraints, called best fit rules that may make it ineligible for forecasting.
Refer to the EXP_SMOOTHING_USE_HIST_AVG_BASE global setting for details on how to specify the initialization of the base forecast.
|