Rough-Tuning and Fine-Tuning 

Use

If a forecast model does not deliver the required results, you can retune the smoothing factors for the following forecast models:

●      First-order exponential smoothing

●      Second-order exponential smoothing

●      Seasonal trend model

Prerequisites

Rough-Tuning

●      You have defined at least one smoothing factor in Customizing.

For more information, see the Implementation Guide (IMG) for Advanced Planning and Optimization under Supply Chain Planning → Service Parts Planning (SPP) → Forecasting → Define Smoothing Factor Set.

●      You have specified one of the defined smoothing factor sets in the forecast profile on the Model Parameter tab page in the Profile Name of Smoothing Factor Set field.

You get to the forecast profile on the SAP Easy Access screen under Advanced Planning and Optimization → Service Parts Planning (SPP) → Planning → Forecasting → Forecast Profile.

●      In the forecast service profile, you have selected either the function Rough-Tuning or the function Composite Forecast and scheduled this service profile in the Planning Service Manager (PSM).

For more information, see Use of the Planning Service Manager in SPP and Planning Services for Forecasting.

Fine-Tuning

●      You have defined the following forecast values in the forecast profile on the Model Parameter tab page:

○       Alpha Factor

○       Increment: Alpha Factor

○       Start Value: Alpha Factor

○       End Value: Alpha Factor

○       Beta Factor

○       Increment: Beta Factor

○       Start Value: Beta Factor

○       End Value: Beta Factor

○       Gamma Factor

○       Increment: Gamma Factor

○       Start Value: Gamma Factor

○       End Value: Gamma Factor

Features

Historical Period for Rough-Tuning and Fine-Tuning

In the forecast profile on the tab page Model Selection you can define which historical period the system considers for rough-tuning and fine-tuning. For the forecast model "First-Order Exponential Smoothing" and for the forecast model "Seasonal Trend Model" you specify the historical period under consideration in the field Periods for First-Order Exponential Smoothing. For the forecast model "Second-Order Exponential Smoothing" you do this in the field Periods for Second-Order Exponential Smoothing.

You get to the forecast profile on the SAP Easy Access screen under Advanced Planning and Optimization → Service Parts Planning (SPP) → Planning → Forecasting.

Rough-Tuning

The system performs rough-tuning in the following cases:

●      You have selected the function Rough-Tuning in the forecast service profile, and have scheduled the forecast service with this forecast service profile as a regular planning service in the PSM.

●      You have selected the function Composite Forecast in the forecast service profile, but the system cannot perform automatic model selection due to stability reasons since the current forecast model has not been used for long enough.

For more information about the exact flow of the composite forecast, see Planning Services for Forecasting

During rough-tuning, the system checks the forecast results for a location product with different smoothing factor combinations as follows:

       1.      It calculates the forecast results for the smoothing factor combinations that contain the selected smoothing factor set. You can define the smoothing factor combinations for smoothing factor sets in Customizing (see prerequisites).

       2.      For further forecast creation, the system selects the smoothing factor combination that has the smallest root of the mean square error (RMSE).

In Customizing, you have defined the following smoothing factor combinations for the selected smoothing factor set:

Alpha: 0.1        Beta: 0.1     Gamma: 0.1

Alpha: 0.1        Beta: 0.2     Gamma: 0.1

Alpha: 0.1        Beta: 0.3     Gamma: 0.1

Alpha: 0.2        Beta: 0.1     Gamma: 0.1

Alpha: 0.1        Beta: 0.1     Gamma: 0.2

For each of these combinations, the system then calculates forecast values and selects the combination with the smallest RMSE for the future forecast creation.

Fine-Tuning

You can only perform fine-tuning manually by selecting the Fine-Tuning pushbutton on the Interactive Forecasting screen. You get to the Interactive Forecasting screen on the SAP Easy Access screen under Service Parts Planning (SPP) → Planning → Forecasting. Fine-tuning can, for example, be useful if rough-tuning did not lead to better forecast results.

During fine-tuning, the system optimizes the alpha, beta, and gamma smoothing factors. To do so, it checks which value delivers the smallest RMSE for each smoothing factor. For each of the smoothing factors, it starts with the value that you have specified as its start value. From this start value, the system continues incrementally by the increment value, which you have specified for each smoothing factor, until it reaches the factor that you have specified as the end value of the corresponding smoothing factor. For each smoothing factor, the system selects the optimal value and writes it in the forecast profile.

Note that all possible values of a smoothing factor are multiplied by all possible values of the other smoothing factors (permutation). For performance reasons, you must therefore not define an increment value that is too small nor too great a distance between the start and end value.

For the alpha factor, you have specified 0.1 as the start value, 0.5 as the end value, and 0.1 as the increment value. The system then checks the forecast result for the following alpha factors:

0,1      0,2        0,3        0,4        0,5

The system enters the value calculated as optimal in the forecast profile in the Alpha Factor field.