Forecast calculation is a method for automatically recalculating the latest period’s actual demand to forecasts for future periods. The number of periods to be calculated is specified in the parameter ‘Extrapolation periods’ in the forecast method.
The forecast calculation flow is illustrated below. It also includes the forecast checks performed as a part of the calculation. If a forecast alarm is activated, a base forecast is selected for the rest of the calculation according to the forecast error logic of the particular forecast method.
The calculation of the forecast is regulated by parameters set for the selected forecast method, and is performed according to the stages below.
The past period’s actual demand, Dr(i), is corrected with the manual forecast adjustment entered. Nonrepresentative demand is not included.
The corrected demand is adjusted for seasonal variations and number of workdays in the period, as shown by:D(i) = Dr(i) / S(i) / P(i).The resulting value is the period’s base demand.
From the resulting base demand, the next period’s base forecast is calculated: F(i + 1). This is done according to the selected forecast formula for the forecast method or forecast simulation methodology.
The forecast is calculated for each of the subsequent periods by seasonally adjusting the base forecast. This is done by multiplying it and the seasonal index for the period and adjusting the forecast for the period length variations by multiplying it and the period length factor, as illustrated below:Fs(i + 1) = F(i + 1) * S(i + 1) * P(i + 1)Fs(i + 2) = F(i + 1) * S(i + 2) * P(i + 2)Fs(i + 3) = F(i + 1) * S(i + 3) * P(i + 3)Etc.
The calculated forecast is adjusted with regards to existing trends. If a trend factor is used for adjustment, then a trend quantity is calculated for each period using the following formulas.Tq(i + 1) = Tf(i + 1) * F(i + 1)Tq(i + 2) = Tf(i + 2) * (1 + Tf(i + 1)) * F(i + 1)Tq(i + 3) = Tf(i + 3) * (1 + Tf(i + 1)) * (1 + Tf(i + 2)) * F(i + 1)Etc.When the trend factors are calculated, the selected trend dampening method is considered, as well as the fact that the factor may not fall below the specified minimum factor or exceed the maximum factor.If a trend quantity is used for adjustment, the trend quantities are calculated for each period using the following formulas.Tq(i + 1) = Tq(i + 1)Tq(i + 2) = 2 * Tq(i + 1)Tq(i + 3) = 3 * Tq(i + 1)Etc.The parameter Forecast with trend (in the forecast method) indicates if the calculated trend quantity should also be added to/subtracted from the forecast of the first forecasted period.The forecast adjusted for season, period length and trend are set to:Fst(i + 1) = Fs(i+1) + Tq(i + 1)Fst(i + 2) = Fs(i+2) + Tq(i + 2)Fst(i + 3) = Fs(i+3) + Tq(i + 3)Etc.
The final calculated forecast is obtained by adding (or subtracting) the manual forecast adjustment to the season-, period length- and trend-adjusted forecast. Ftot(i + 1) = Fst(i+1) + M(i + 1)Ftot(i + 2) = Fst(i+3) + M(i + 2)Ftot(i + 3) = Fst(i+3) + M(i + 3)Etc.
|D(i)||=||Base demand during period (i)|
|Dr(i)||=||Actual demand during period (i)|
|F(i)||=||Base forecast for period (i)|
|Fs(i)||=||Seasonally adjusted and period-length adjusted base demand for period (i)|
|Fst(i)||=||Base demand adjusted for season, period length, and trend for period (i)|
|Ftot(i)||=||Total forecast for period (i)|
|M(i)||=||Manual forecast adjustment for period (i)|
|P(i)||=||Period length adjustment factor|
|S(i)||=||Seasonal index for period (i)|
|Tf(i)||=||Trend factor for period (i)|
|Tq(i)||=||Trend quantity for period (i)|