Mathematical Forecasts

Use mathematical forecasts to calculate a forecast based on the forecast history and different kinds of demand, using user-defined values. By defining the number of periods to use from the forecast history, the forecast is calculated using a forecast formula. The forecast formula defines how the calculation of the forecast is done.

The kind of demand controls what to calculate forecast on. In order to control the forecast calculated and to get it to reflect actual demands, a number of settings may be used.

Before you start

The following settings must be met in order to use mathematical forecasts:

  • In 'Forecast Method. Open' (FCS300), set the Forecast formula to 1 – 4 or 11 – 14. Depending on the formula selected, enter values for the following:
  • Seasonal influence
  • Alpha factor
  • Beta factor
  • F/C with trend
  • Extrapol w trend

Description

Let's start with a moving average method, we use formula 1 (moving average) together with 2 calculating periods. We want to update the annual demand and we are using alarms and saving them but no change to the forecast if alarms are activated.

We extrapolate 12 months in the future no trend or seasonal influence, but we want to calculate MAD and ME. 12 periods are kept in the forecast history and 6 periods of future forecast (extrapolated values) are kept in item statistics. The annual demand entered in 'Item. Connect Warehouse' (MMS002) is used for calculations.

We start the calculation by running 'Automatic Forecast. Calculate' (FCS100). The result in 'Automatic Forecast. Open' (FCS001) will be as follows the base forecast, which is the forecast evenly spread over the year no consideration of seasonal influence. In this case the annual demand 1200 is divided by 12 and the forecast for period 0212 is 100, we extrapolate 12 months in the future with no trend or seasonal influence, 100 per period.

By entering a period and setting a demand it is possible to simulate that time passes. I will enter the demand for 0212 as 90 and the demand for 0301 as 120 and then calculate the forecast for 0302 by running 'Automatic Forecast. Calculate' (FCS100) for period 9902-9902. This is best done by pressing F20 in 'Automatic Forecast. Open' (FCS001) or by running 'Automatic Forecast. Calculate' (FCS100).

This gives us an average over 2 periods of 105 ((90+120)/2), this is my new forecast and my extrapolated value. The MAD is calculated as ( |120-95| + |90-100|)/2 = 17.5 Mean Errors is ((120-95) +(90-100))/2 = 7.5. All calculated values are found in 'Automatic Forecast. Open' (FCS001), on panels E and F. You go to the F-panel by page down.

The function-keys in 'Automatic Forecast. Open' (FCS001) are very useful: F15=Manual adjustment, F16=Simulation values, F17=F/C alarms, F18=Manual F/C, F19=Display graphic and F20=Calculate F/C.

The alarm 2 has been activated because the difference between demand and forecast is too great. This can be viewed in 'Alarm Item. Display' (FCS010), which can be reached by F17 in 'Automatic Forecast. Open' (FCS001)

Every time we run 'Automatic Forecast. Open' (FCS100), it’s possible to calculate some planing values and update the yearly demand, but the method for each value must not be set to zero.

If we look in 'Item. Connect Warehouse' (MMS002) the yearly demand is updated to 1245. (100+95+10*105)

Related topics