Defining ATT forecast engine

Select Configuration > Cycle > Engines.

Click New and select Forecast Engine (ATT).

Specify this information:

Name: The name of the forecast engine.
Display Name: The name displayed in the application.
Users: The users authorized to access the forecast engine.
Roles: The roles of the user(s) authorized to access the forecast engine.
Type: The forecast engine type. This is a read-only field based on the selection specified when defining a new forecast engine.
Module: The cycle or cycle and module for which the forecast engine must be used.
Tags: The tag linked to a forecast engine. Select from a list of pre-defined tags displayed in the window. You can link one or more tags to a forecast engine. This option is used to filter the type of forecast engines that must be displayed.

Specify this information on the Default Settings tab:

Default Algorithm

The default algorithm used for the forecast engine. Possible values:

Best
Crostons
Holt-Winters
Least Squares
Moving Average
Exponential Moving Average
Multiple Regression
Bayesian Analysis of Time Series
Events only
Naive
Arima

SMP Measure

The measure that contain the values, to indicate the item and location combinations which are considered as Slow Moving Products (SMPs).

Spreading Measure for period constant results

The measure that must be used when spreading the forecast engine output to item, location and pconst.

Spreading Measure for time-phased results

The measure that must be used when spreading the forecast engine result to item, location and period.

Perform Event Modeling

Select this option to apply event modeling. This enables the forecast engine to use the defined events and execute event modeling, in addition to the selected algorithm.

If this option is set to On, the Event Profile measure on the Measure Mapping tab must be selected to store the event profile. Optionally, the Average Event Size measure can be selected to store the average event size statistics.

Item Level

The default item level at which the forecasts must be generated.

Location Level

The default location level at which the forecasts must be generated.

Period Level

The default period level at which the forecasts must be generated.

Forecast Combining

Indicates, if the engine implements Forecast Combining.

Note:

This option is enabled only if the Default Algorithm is set to Best.
If this option is set to On, the Forecast Combining tab is enabled.

Specify this information, if applicable, on the Forecast Engine tab:

Classic Combining: If this option is set to On, the engine combines all the generated forecast and provide the best forecast based on varied methods such as Simple Mean, Trimmed Mean, Winsorized Mean, Weighted AIC).
Machine Learning Train-test: If this option is set to On, the train-test data is used for the MLR run.
Machine Learning Train-test randomization: If this option is set to On, a combined test with the random number seed run MLR using the train-test methodology on the randomized data set.
Machine Learning Cross-Validation: If this option is set to On, the cross-validation data split method is used for the MLR run.
Machine Learning Full Dataset: If this option is set to On, the full data set is used for the MLR run.
Machine Learning Full Dataset randomization: If this option is set to On, a combined test with the random number seed run MLR on the full randomized data set.

Click New on the Measure Mappings tab and specify the required information.

Note: On the Parameters tab, a list of parameters, that are used by the forecast engine, to generate a forecast are displayed. If the parameter values are not defined, the application does not consider these parameters.

Specify the required information on the Independent Variables (used to calculate a regression analysis against the dependent variable) tab,

Note:

This tab is enabled only if you select the Multiple Regression in the Default Algorithm field.