Creating Auto Predictions for Forecasts

Create an Auto Predict prediction definition to automate predictions.

To create an Auto Predict prediction definition:

  1. From the Home page, click Application, then click Overview, and then from the Actions menu, select Auto Predict.
  2. On the Auto Predict page, click Create.
  3. Enter a name and a description for the Auto Predict definition.
  4. From the Select Cube list, select the cube that contains the historical data. This cube is also used to store prediction results. If you want to store prediction results in a cube other than the cube with the historical data, select Cube to Cube.

    If you select Cube to Cube, select a Source cube (the cube that has the historical data you want to analyze) and a Prediction cube (the cube where you want to store the prediction results). The Source cube is typically an ASO cube (although BSO cubes are supported). The Prediction cube can be an ASO cube or a BSO cube.

    Note:

    If you are analyzing and predicting in the same cube, you can store prediction results in a different scenario/version.
  5. Define the slice of data for analysis (where the historical data is), and then define the slice of data to store the prediction results:
    • In the Analyze section, define the slice of data for the historical data source by selecting members for each dimension.
    • In the Predict section, define the slice of data to store the prediction results for the base prediction by selecting members for each dimension.

    Tips for selecting members:

    • Click a member in the Analyze section to open the Member Selector.
    • You can select members for each mapped dimension by selecting the dimension from the Dimension list in the Member Selector.
    • After selecting members for mapped dimensions, click OK. Then click a member of an unmapped dimension and select members for each unmapped dimension.
    • For the Analyze section, you can select single members or multiple members. You can select by member, or select by member functions or substitution variables.
    • In the Predict section, members are pre-selected with the members selected in the Analyze section. Click a member in the Predict section to open the Member Selector if you need to change the selection. For example, for the Scenario dimension, change the member from Actual to Forecast.
    • In the Predict section, you can change the member selections only if there is a single member selected in the Analyze area. Multiple member selections are common to both Analyze and Predict.
    • Unmapped dimensions can be defined using single members only.
    • Select members using these guidelines. Predictions using other combinations would require writing results to read-only cells or take too long:

      • In the Analyze section, where historical data is, select members with these combinations of data storage and cube type:

        • BSO—Store and Shared members

        • ASO—Any member without a member formula, and for account dimension, only leaf level members (level 0 members)

      • In the Predict section, where you want to perform predictions, select members with these combinations of data storage and cube type:

        BSO or ASO—Store/Never Share

  6. In the Analyze section, specify how much historical data you have. Click Last 3 Years, the default, and slide the slider to the number of years of historical data you want to use in the prediction, up to a maximum of five years.
  7. In the Predict section, specify how far you want prediction results to be calculated. Click Select Start Period and choose a method:
    • Slide the slider to the number of periods you want to predict, up to a maximum of 48 periods. The default is six periods.
    • To define a fixed date range, click More Options, and then select the Start Period, End Period, and Number of Periods.
    • To define a scenario-based date range, click More Options and then click Use a scenario-based date range. Then select the Scenario to use, the End, and the Number of Periods. For End, you can select a fixed period date for the end of the predictions, or base it on the end date of the selected scenario. Using a scenario-based date range restricts prediction calculations to the range defined for the scenario time period. For example, you can predict based on the start of the forecasting periods from the Forecast scenario.

    You can predict at a weekly level in Financials. To predict at a weekly level, when defining the Auto Predict definition, in the Predict section, you must select Specific Period for the Start Period.

    Tip:

    When setting the Start for a prediction, when the start is Specific Period, for the Select Period Member Selector, you can select by substitution variable for the Current Period. In this way, job definitions are updated each time the value for the substitution variable is updated—change the value for the substitution variable and you don't have to edit every job to update the current period. To select for Substitution Variables, in the lower left corner of the Member Selector, select Substitution Variable from the list.
  8. Optionally, to include Best Case and Worst Case results in the prediction results, click Add scenario icon on the right side of the Predict area and then click Best Case and Worst Case. Then select members for the dimensions. Note that you can select members only for dimensions with single member to single member mapping. Dimensions with multiple member selections and unmapped dimensions use the same selections for Best Case and Worst Case prediction results.

    Tip:

    You can remove Base Prediction, Best Case, or Worst Case from the prediction definition. Click ActionsEllipses next to the column you want to remove, and then click Remove.

    The prediction interval used to define the Best Case and Worst Case results is defined in Settings for the application and is shared with Predictive Planning. See What Application and System Settings Can I Specify?.

  9. Click Save to save the prediction definition.

After defining the Auto Predict prediction, you can run the prediction from the Auto Predict page, or you can schedule a job to run the prediction automatically. See Running Auto Predict Predictions.

Auto Predict definitions are saved as global artifacts and are backed up with the maintenance snapshot.