Forecast Settings
If you have run any pipelines, they will need to be run again after changing these settings.
CAUTION
Changing any of these project settings in an external forecast project can lead to errors.
- Number of forecast periods (horizon)
-
Specify the number of periods to (also called the ). For example, if the time interval for this project is set to WEEK, specify 12 to generate 12 weeks of forecasts for the dependent variable.
- Number of periods to exclude from modeling
-
Specify a positive integer for the number of time periods to leave out when fitting models. The number of periods to exclude from modeling is also known as the out-of-sample region.
This is used to determine performance statistics for the models selected for each . Removing the most recent time periods provides a better assessment on how well the model generalizes, or how well it performs on data that was not used to fit or select the model.
Forecasts start in the out-of-sample region. If the out-of-sample region is greater than or equal to the horizon, no forecasts are made in the future periods after the end of the historical data. For more information, see Using an Out-of-Sample Region.
If the project has been imported or upgraded from previous versions of SAS Visual Forecasting, it might not show any statistics for an out-of-sample region, if one is specified. For more information, see Imported Project Does Not Show Out-of-Sample Region.
Note: If you create any based on out-of-sample forecast attributes, those filters are disabled if you change this setting back to 0. - Confidence limit
-
Specify the size of the confidence level for the forecasts. By default, this confidence level is 0.05, which is a 95% confidence limit.
- Champion selection criteria
-
Choose the statistics of fit analysis that is used to determine the champion pipeline for this project. The statistics of fit are statistical values that are used to evaluate how well a forecasting model performs by comparing the actual data to the predictions. For a given forecast model that has been fitted to the time series data, the model should be checked or evaluated to see how well it fits or forecasts the data.
Each modeling node in a pipeline has its own model selection setting to determine the best model to use from within that node. The Champion selection criteria is used to select the best pipeline when more than one pipeline is run.
For a full description of the statistics of fit options that are available, see Model Selection Criteria.
- Allow negative values for forecasts and overrides
-
Select this option to allow negative values for forecasts and overrides. If you clear this check box, then any negative values in the forecast model are set to 0 and any overrides with negative values are removed.
This setting is turned off by default for time series projects. For external forecasts, this setting is turned on by default.
If Missing interpretation for the dependent variable is set to 0, you could still see some negative values for the lower (Forecast LCL) in the project time plots.
Changes to this setting invalidate any existing overrides that have been created. Changes also invalidate the pipelines.
Note: To allow negative overrides, you must also set the Maximum decrease percentage on the Overrides tab to greater than 100%.