Forecast Engine Combining added trim factor parameter

The forecast engine configuration has a new, optional parameter:

  • MLR Trim Factor

MLR = Multivariate Linear Regression (MLR) and relates to the algorithm used for Machine Learning techniques in forecast combining added in a previous version.

The default value for the parameter is null, meaning no trimming takes place. Setting a value, removes the highest “x” and lowest “x” forecast in each time period prior to the combining process. This is a one-sided value, i.e. when x = 1 it removes the upper extreme value and the lower extreme value.

This is applicable where the Default Algorithm = Best and Forecast Combining = On. The trim factor is applied for any of these combining methods, when active:

  • ML Train-Test
  • ML Train-Test randomization
  • ML Cross-Validation
  • ML full dataset
  • ML full dataset randomization

All forecast engine definitions, in the standard Demand Planning content, do not have a value set. Therefore, the behavior is based on the previous versions.

Note: This feature is enabled by adding or editing a forecast engine and setting the MLR Trim Factor parameter to a positive integer value and no additional role or privilege is required to access this feature.