Overview

The most likely problem with your initial forecasta numerical prediction of a future value for a specified time period for each unique combination of BY variable values model is that the model ignores a known eventan incident that disrupts the normal flow of any process that generates the time series. Examples of events are holidays, retail promotions, and natural disasters.. You can improve the accuracy of your model by adding events. When you add events in a project, SAS Visual Forecasting must reforecast the project in order for the models to reflect these changes.

An event is an incident that disrupts the normal flow of a process that generates the time seriesan aggregation of transactional data into specified time intervals and sorted according to unique combinations of the default attributes (BY variables). Examples of events are holidays, retail promotions, and natural disasters. Defining an event enables you to model the effect that special events have on the dependent time series. When you apply an event to a time series, SAS Visual Forecasting creates an indicator variable. This variable indicates the occurrence of the event at the time period for which the event is defined. This indicator variable is used as a regressor variable for time series modeling and forecasting.

For example, retail sales data follows a fairly steady pattern depending only on the day of the week. However, when you include events such as a one-time New Year's Day promotion, the forecasting model can predict a temporary increase in sales and then return to normal sales levels after the event. Some yearly events that occur on a fixed date, such as New Year's Day, can be modeled as part of the regular seasonal model. Seasonal events that are most effectively modeled as regressors are those that occur on a different date each year, such as Thanksgiving.

You can use SAS Visual Forecasting to predict the effect of a future event. Before the forecasting model can predict the effects of a future event, your data must include the effects of past occurrences of the event. For example, after reviewing your December sales data in the United States, you decide to create an event for next Christmas. SAS Visual Forecasting must have data that explains how retail sales were impacted by previous Christmas events. Without this data, SAS Visual Forecasting cannot predict the effect of the next Christmas event on retail sales.

SAS Visual Forecasting includes a list of predefined events. You can also provide your own events.

• To create events in the project, see Creating Events for a SAS Visual Forecasting Project.
• You can create an event definition table as described in Creating Event Definitions.

When working with events, start on the Data tab and select Events in the left pane. For a new project, Predefined events and Custom events are listed. Predefined events includes events for each holiday. After creating your own events, you can see them listed under custom events. Any events imported from an event definition table are listed by the table name.

The following tools for working with events are shown in the toolbar in the middle pane.

Opens the New Event window to create a new event. The new event is available for any modeling node or time series within this project. After the event is saved, it is added to the list of custom events.

Opens the Edit Event window for a selected event. Only events created in the project can be edited. Predefined events and events imported from an event definition table cannot be edited.

Opens the Copy Event window with the settings for the selected event. Use this window to create a new event based on the event that you have selected.

See Also

• Naive Model
• Customizing Pipelines for Each Segment