Ansicht
Dokumentation

APO_SPP_TEST_SEQ - Configure Automatic Model Selection

In this IMG activity, you can make the following settings for automatic model selection:

1. Sequence of forecast tests
2. Forecast strategies
3. Weighting factors for forecast errors

1. You can define the sequence in which the forecast tests (such as, intermittent forecast test, seasonal test, trend test) are executed in the "Test Sequence" field. If a test is successful, the system ends the series of tests, that is, no further tests are carried out. Note that the "initial" forecast test type is permitted; this is always used if none of the forecast tests are successful. We therefore recommend that you always assign the highest test sequence number to the "initial" forecast test type.
2. You can assign forecast strategies (FS) to every forecast test. If a forecast test is successful, all the assigned forecast strategies will run. The forecast strategy that results in the minimum number of errors is adopted by the system as the optimum in the forecast profile. After this, the system terminates optimization.
3. Using the forecast model weighting factor, the error calculated for the forecast model can be weighted. A higher value gives rise to a higher weighting of the error - the corresponding forecast strategy is determined as optimum "less often" during automatic model selection