Show TOC
## Automatic Adaptation of the Alpha
Factor

### Use

### Features

### Activities

##### Forecast Strategy 12

##### Forecast
Strategy 23

This model is used in forecast strategies 12 and 23. The system automatically optimizes the alpha factor.

In forecast strategy 12, the alpha factor is changed within a specified range using a specified increment size. A forecast is carried out for each alpha value. The system then chooses the value for which the chosen the measure of error is the smallest.

In forecast strategy 23, the alpha factor is adapted in every ex-post period based on the mean absolute deviation (MAD) and the error total (ET) as in the formula below.

The minimum alpha factor is 0.05. The maximum alpha factor is 0.90.

In the univariate forecast profile you specify the measure of error to be used, the default is MAD - Mean Absolute Deviation. You can also enter the start and end value for alpha, together with a value for the increment. If you do not make any entries, the system uses default values of start value 0.1, end value 0.5, increment 0.1.

The sytem selects the value of alpha with the lowest measure of error.

The minimum alpha factor is 0.05. The maximum alpha factor is 0.90.

The initial alpha factor
(a _{0}) is either the
default of 0.3, or a value that you have entered in the univariate forecast
profile.

The system uses the following iterative formula to calculate the alpha factor in each ex-post period:

- TS is the tracking signal.
- TS = 0 if MAD(i) = 0.
- TS = ABS [ET (i) / MAD (i) ] if MAD (i) ยน 0.
- i = 1 … n, where n is the number of periods in the ex-post forecast.
- a
_{i}is used to calculate the ex-post forecast in the period i+1.