Trend in Forecast Methods
You can add trends to a forecast calculation in the Forecast Method Property dialog box.
- To access Forecast Method Property, select .
- Click or .
- Select Property Trend and select the applicable Trend Methods from the drop-down list.
- Depending on the selected Trend, the next steps are optional.
- If required, specify the Trend Damping to reduce the trend factor period by period when extrapolating forecasts using a trend.
- Specify the number of catch up periods.
- Select the beta factor for smoothing.
- Select the Add trend to first calculated period check box if the defined trend should be added to the first calculated period of the forecast.
Trend constant
The forecast is extrapolated with a constant trend factor. You can select if this trend factor is added in the calculation period. This is done by selecting Add trend to first calculated period.
Trend factor preceding period
The trend factor is calculated according to:
Where:
-
= Trend index for period i - n = Trend period length
- i = Period i
Trend factor is calculated using demand history and depends on trend period length. This is referred to as n in the formula. When calculating trend factor, the total demand for a number of periods is compared to the total demand for an equal number of preceding periods. The calculated trend factor is compared to the minimum and maximum allowed value of the trend factor. If the calculated trend factor is greater than the trend factor max, the trend factor max is used and vice versa. You can also select if this trend factor is added in the calculation period. This is done by selecting Add trend to first calculation period.
If the trend period length is three forecast periods, the trend index calculation is based on the ratio of the demand from the three immediately preceding forecast periods (1, 2, and 3 periods ago) and demand from the three forecast periods before that (4, 5, and 6 periods ago).
The extrapolated forecast is calculated according to:
Where
is the base forecast with or with out the trend (depends on if
Add trend to first calculated period is selected or not).
Trend factor preceding year
The trend factor is calculated according to:
Where
-
= Trend index for period i
- n = Trend period length
- i = Period i
- y = One year back in time
Trend factor is calculated using demand history and is depending on Trend period length. This is referred to as n in the formula. When calculating trend factor, the total demand for a number of periods is compared to the total demand for the same n periods the previous year. The calculated trend factor is compared to the minimum and maximum allowed value of the trend factor. If the calculated trend factor is greater than the Trend factor max, the Trend factor max is used and vice versa. You can also select if this trend factor is added in the calculation period. This is done by selecting Add trend to first calculation period.
The extrapolated forecast is calculated according to:
Where
is the base forecast with or with out the trend (depends on if
Add trend to first calculated period is selected or not).
Trend year by year
This trend is commonly called trending as the average of previous years. This method calculates for a full year and considers the number of weeks for the present and following years.
You can use this method with these options:
- Apply the trend from the start calculation point.
- Lock the trend calculation to the year
borders.
In this option, the trend remains constant for the entire year and calculated by using data from the previous true years. You can use this option in calculating forecast with seasonal influences.
Trend quantity
Trend quantity is calculated according to:
where
-
= Calculated trend
quantity
- β = Smoothing constant used to control how fast the trend is updating the forecast.
When working with trend quantity, it always looks two periods back
when
.
The trend quantity indicates the difference between the average demands in two consecutive forecast periods. It is calculated automatically using exponential smoothing. The beta factor is used as a smoothing constant for this calculation. When extrapolating forecasts for future periods, the forecast value for the period is calculated as the forecast value from the preceding period plus the trend quantity.
Trend measure
The trend factor is defined in an assigned measure (data type, budget). The extrapolated forecast is calculated according to:
You can assign a measure in these ways:
- Through the dataset properties
- As a data type + budget
- As a calculated measure name
Trend damping
You can damp the influence of the Trend Factor Preceding Periods and Trend Factor Preceding Year when extrapolating forecasts. The trend dampening is used to reduce the trend factor period by period when extrapolating forecasts using a trend. In this way, you can reduce the influence of trend on a forecast in proportion to increase future uncertainty.
Two damping methods are supported:
- The trend factor for the first period is equal to the calculated trend factor. The trend factor for each subsequent period is obtained by multiplying the previous periods trend factor by the calculated trend factor.
- The trend factor for the first period is equal to the calculated trend factor. The trend factor for each subsequent period is obtained by multiplying the previous periods trend factor by the β factor.
Damping method
- Formula = 3 periods moving average
- Extrapolation periods = 3
- Trend period length (TREP) = 2
- Trend Factor Min = -0,5
- Trend Factor Max = 1
- Damping method 1
The trend factor for each subsequent period is obtained by multiplying the previous period's trend factor by the calculated trend factor.