Creating forecast engine

  1. Select Global > Business Area > Configuration > Forecast Engines. The Forecast Engines screen is displayed.
  2. Click New. The Engine Details tab is displayed.
  3. Specify this information on the Engine Details tab:
    Name
    A unique name of the forecast engine.
    Display Name
    The display name of the forecast engine.
    Type
    The type of forecast engine.
    Note: You can set this field to ATT during the creation of forecast engine. You cannot modify the value after the forecast engine is created.
    Algorithm
    The default algorithm used by the forecast engine. Possible values:
    • Best
    • ARIMA
    • BATS
    • Crostons
    • Exponential Moving Average
    • Holt-Winters
    • Least Squares
    • Moving Average
    • Multiple Regression
    • Naive
    Event Modeling Method
    The event modeling method of the forecast engine.
    Note: By default, this field is set to None.
    Optimize Parameters
    Indicates if the parameters are optimized in the forecast engine. Possible values:
    • On: Optimization of parameters occurs in the forecast engine before generating forecasts.
    • Off: The specified parameter values or engine default values are considered for generating forecast.
    Note: 
    • By default, this field is set to Off.
    • This switch is enabled, only if the value in the Algorithm field is set to Best, BATS, or Holt-Winters.
    Forecast Combining
    Indicates if the forecast engine considers Forecast Combining as part of Best fit.
    Note: 
    • By default, the value is set to Off.
    • This switch is enabled only if the value in the Algorithm field is set to Best.
  4. Specify this information on the Forecast Combining tab if enabled:
    Classic Combining
    Indicates if the engine combines all the generated forecasts to create a best forecast based on various methods such as Simple Mean, Trimmer Mean, Winsorized Mean or Weighted AIC.
    Machine Learning Train-test
    Indicates if the train test method is used to execute the MLR run.
    Machine Learning Train-test Randomization
    Indicates where to use the train test method with the random number seed to execute the MLR run.
    Machine Learning Cross-Validation
    Indicates if the cross-validation data split method is used to execute the MLR run.
    Machine Learning Full Dataset
    Indicates if a combined test MLR run is executed with the full data set.
    Machine Learning Full Dataset Randomization
    Indicates if a combined test MLR run is executed using the full data set with the random number seed.
    Note: The Forecast Combining switch on the Engine Details tab must be enabled to modify the Forecast Combining tab.
  5. Specify the value of the required parameters on the Parameters tab.
    Note: The parameters are displayed based on the value selected in the Algorithm field of the Engine Details tab. See Forecast engine parameters, for more details.
  6. Click Save.
Note: On this screen, you can:
  • Delete forecast engines using the Delete option. See, Deleting forecast engine, for more details.
  • Modify an existing forecast engine using the Edit option.
  • Create a duplicate of the existing forecast engine and modify the parameters as required using the Duplicate option.