Forecasting Parameters for Hyperparameter Tuning

This topic lists the forecasting parameters that you can add to your user-defined forecasting profile of the Bayesian machine-learning type to control hyperparameter tuning for your demand or demand and supply plan.

This table lists the forecasting parameters in the Hyperparameter Tuning category for hyperparameter tuning:

Name	Description
HypertuneAdvisor	Specifies whether to provide notifications through the Planning Advisor when hyperparameter tuning is successful for a forecasting node (combination). When the value is 0 (zero), notifications aren't provided through the Planning Advisor. When the value is 1, notifications are provided through the Planning Advisor. Default: 0
HypertuneAdvisorPageLayout	Specifies the name of the page that's linked to the notification of the Summary of hyperparameter tuning type in the Planning Advisor. This forecasting parameter should be used when the HypertuneAdvisor forecasting parameter is set to 1. When the HypertuneAdvisorPageLayout forecasting parameter is used, the View More button is available in the details of the hyperparameter-tuning notification. Default: null
HypertuneDisableSlowMethods	Specifies the forecasting methods that should be turned off during hyperparameter tuning for a forecasting node. However, the disabled forecasting methods are used during the final forecast for the forecasting node. The forecasting methods are referred to by their letters. When the value is null, the value is read as CEK, which means that the Multiplicative Monte Carlo Regression (C), Combined Transformation (E), and Multiplicative Monte Carlo Intermittent (K) forecasting methods are disabled. When the value is 0 (zero), no forecasting method is disabled. Default: CEK
HypertuneHalflife	Specifies the number of days for which the results of hyperparameter tuning for a forecasting node are considered as fresh and for which hyperparameter tuning won't be attempted again for the forecasting node. Default: 182
HypertuneHoldback	Specifies the number of days before the plan start date for holdback validation (out-of-sample testing) for a forecasting node. This forecasting parameter isn't used if its value is greater than the days reflected by the value of the HypertuneHoldbackPercent forecasting parameter. The forecasting nodes are selected for these days according to the value of the HypertuneSamplePercent forecasting parameter. Default: 182
HypertuneHoldbackPercent	Specifies the maximum percentage of history before the plan start date for holdback validation (out-of-sample testing) for a forecasting node. This forecasting parameter is used if the resulting days are less than the value of the HypertuneHoldback forecasting parameter. The forecasting nodes are selected for this period according to the value of the HypertuneSamplePercent forecasting parameter. Default: 30
HypertuneMAPEThreshold	Specifies the minimum difference required between the values of the tuned base mean absolute percentage error (MAPE) and tuned best MAPE for a forecasting node. When the tuned best MAPE is less than the tuned base MAPE, and the difference is equal to or more than the value of this forecasting parameter, hyperparameter tuning for the forecasting node is considered as successful. Otherwise, hyperparameter tuning is ignored for the forecasting node. Default: 1
HypertuneMaxPermutations	Specifies the maximum number of permutations of hyperparameter-tuning values to be evaluated for a forecasting node through a random-walk algorithm. If the value is 0 (zero) or 1, all permutations are considered for hyperparameter tuning. Any value greater than 1 is the maximum number of permutations that are evaluated for the forecasting node through the random-walk algorithm. For details on when the forecasting-parameter value is used and the random-walk algorithm, see the topic titled How You Improve the Performance of Hyperparameter Tuning. Default: 100
HypertuneOutputMode	Specifies how the results of hyperparameter tuning should be saved for a forecasting node. If the value is 0 (zero), the base forecast for the forecasting node is saved in the output measure of the forecasting profile. If the value is 1, the tuned forecast is saved in the output measure of the forecasting profile if the difference between the base MAPE and best MAPE equals or exceeds the value of the HypertuneMAPEThreshold forecasting parameter. Otherwise, the forecast according to the forecasting profile's forecasting parameters is stored in the output measure. The tuned settings are stored in the Bookings Forecast: Tuned Settings, Shipments Forecast: Tuned Settings, or <user-defined output measure>: Tuned Settings measure. If the value is 2, the tuned forecast is stored in the Bookings Forecast: Tuned, Shipments Forecast: Tuned, or <user-defined output measure>: Tuned measure so long as the difference between the base MAPE and best MAPE equals or exceeds the value of the HypertuneMAPEThreshold forecasting parameter. The forecast according to the forecasting profile's forecasting parameters is stored in the output measure. The tuned settings are stored in the Bookings Forecast: Tuned Settings, Shipments Forecast: Tuned Settings, or <user-defined output measure>: Tuned Settings measure. Default: 2
HypertuneParamSet	Specifies the default set of parameters for hyperparameter tuning of the forecasting parameters. This set is used when hyperparameter tuning is enabled, but no values are provided in the Hyperparameter Tuning Values column in the Parameters step in the guided process for creating or editing a forecasting profile. The default set is as follows: IntermitCriterion[10,90],ForecastValidationSensitivity[0.2,5,10],OutlierSensitivity[2,3,4],CollinearityUseRidge[0,1] Note: Don't change the value for this forecasting parameter.
HypertuneSamplePercent	Specifies the percentage of the forecasting nodes for which hyperparameter tuning should be attempted in a plan run. If the value is 0 (zero), hyperparameter tuning is turned off. If the value is 1 or more, hyperparameter tuning is turned on. If the value is 100, hyperparameter tuning is performed for all the forecasting nodes in the plan. The forecasting-parameter value must be 0 or a positive integer. Default: 0

This table lists the forecasting parameters in the Nodal Tuning category for hyperparameter tuning:

Name	Description
EnableNodalTuning	Specifies whether to use the tuned settings for a forecasting node for which hyperparameter tuning was successful in a plan run as the forecasting-parameter settings in later plan runs until the value of the HypertuneHalflife forecasting parameter is reached. The values can be 0 (zero), 1, or 2. If the value is 0 (zero), nodal tuning isn't used. For information on how to use this forecasting parameter, see the topic titled How You Enable Nodal Tuning for Forecasting Nodes. Default: 2
SetNodalTuningSource	Specifies the measure that has the settings for nodal tuning. If the value is 0 (zero), the Bookings Forecast: Tuned Settings, Shipments Forecast: Tuned Settings, or <user-defined output measure>: Tuned Settings measure is used. If the value is 1, a selected, user-defined measure is used. This forecasting parameter can be used only if the value of the EnableNodalTuning forecasting parameter is 1 or 2. Default: 0 Note: This forecasting parameter isn't presently supported and shouldn't be used.