Use
If you do not want to specify a forecast model manually, you must instruct the system to make an automatic selection. With automatic selection, the system analyzes the historical data and then selects the most suitable model.
Prerequisites
The system requires a different number of historical values for different tests. For further information, see
Model Initialization.Features
If you want to select a model manually, you must first analyze the historical data to determine whether a distinct pattern or trend exists. You then define your forecast model accordingly.
The following models are possible:
If the system cannot detect any regular pattern in the historical data, it automatically selects the constant model.
Model Selection Procedures
You choose between two procedures for automatic model selection:
Procedure 1
If you want the system to select the forecast model, you can choose between various statistical tests and test combinations which determine the model. The tests that are carried out depend on how much you know about the past data pattern.
Trend test
In the trend test, the system carries out a regression analysis on the historical values and finds out if there is any significant trend.
Seasonal test
In the seasonal test, the system ignores any possible trends in the historical values and then carries out an auto-correlation test.
|
Pattern |
Test |
|
No information |
Test for trend and seasonal patterns |
|
No trend |
Test for seasonal pattern |
|
No season |
Test for trend pattern |
|
Trend |
Test for seasonal pattern |
|
Season |
Test for trend pattern |
If you know that a particular pattern exists or does not exist before the model is selected, you can have the system test the historical time series for this trend or seasonal pattern.
If you are unaware of the historical pattern, the system carries out a trend test and a seasonal test. The forecast model is determined on the basis of which test produces the more significant results.
Procedure 2
The system calculates the models to be tested using various combinations for alpha, beta, and gamma. The smoothing factors are also varied between 0.1 and 0.5 in intervals of 0.1. The system then chooses the model which has the lowest mean absolute deviation (MAD). Procedure 2 is more precise than procedure 1, but takes much longer.
To work with procedure 2, choose the forecast strategy Automatic Model Selection with Process 2 when you create a forecast function.
See also:
Manual Model Selection