Specifying forecast strategy options
Use the Strategy pane of the Forecast editor to specify how the forecasted data is formatted.
To specify strategy options:
About this task
- Open the Model page of the Forecast editor
- Select the column that contains the values to forecast
from the Forecast column drop-down list.
All available numeric columns are displayed in the Forecast column drop-down list.
- Specify how the forecasted values will be aggregated in
the Aggregation strategy drop-down list.
The aggregation strategy is largely dependent on the type of the value column that you are forecasting. In most cases you would use the Group sum option. There are instances however, such as aggregating temperatures, where you would use the Group average option.
- Specify how the forecast model is constructed in the Construction
strategy drop-down list. If you select Leaf, the model is calculated at leaf level and aggregated upwards using a bottom-up approach. This strategy is most effective when there is a full history at the lower reporting level (for example, a rich data set with no nulls at the bottom level). If you select Root, the model is calculated at root level and distributed to leaf nodes using a top-down approach. A root level strategy is more effective when there is a smaller data set.
- If you select the Root construction
strategy, select how to distribute the root level forecast value to
leaf levels in the Distribution strategy drop-down
list. Available values include:
- Equal - Equally distributes the root level forecasts to each leaf level.
- Historical total - Distributes root value forecasts proportionally to sub-levels according to the total contribution of each group across the whole historical data set.
- Most recent period - Distributes root value forecasts to the most recent forecasted period.