Configuring Advanced Predictions

You can use machine learning algorithms to generate advanced multivariate predictions.

If you have not already created a calendar, create a calendar to use with your advanced predictions. You can also configure events to associate with Advanced Prediction prediction definitions. Configure calendars and events first, before you define an advanced prediction. See Configuring Calendars and Events.

Advanced Predictions require you to use a predefined calendar.

1. Enable Advanced Predictions. See Enabling AI in FreeForm.
2. Prepare, review, and modify your input driver data as needed.
3. From the Home page, click IPM and then click Configure.
4. Click Create.
5. In IPM Configurator, on the Types page, enter a name and description.

   The name should be alpha numeric only, without any special characters.

6. Click Advanced Predictions and then click Next.
7. On the Calendar page, select a calendar to use, and then click Next. The Historical Data section and the Future Data section are populated with details from the selected calendar. The cube selection also comes from the calendar definition.
8. On the Historical Data page, define the slice for the prediction, the output to predict, and the drivers to use for input for historical data, and then click Next. This information is used to train the model that drives the predictions. These selections define what you want to predict and what the input drivers are.
   • Define Model Scope. See Defining the Model Scope.
   • Select Output to Predict. See Selecting the Output to Predict.
   • Select Drivers as Input. See Selecting Drivers as Input.
9. On the Future Data page, in the Define Model Scope section, define where to store the prediction results and the data slice for drivers in future periods, and then click Next. See Defining the Model Scope.

   You don't need to make any changes to Select Output to Predict or Select Drivers as Input.

10. On the Prepare Data page, define how to assess the data before performing the prediction, and then click Next. See Preparing the Data.
11. On the Model Settings page, select the algorithm, forecast error method, and confidence intervals to use for the prediction, and specify whether to include events in the prediction, and then click Save.

    Advanced Predictions automatically detect seasonality and perform data screening.

    • Select Algorithms—Select the algorithms to use for the advanced prediction. See Selecting Algorithms.
    • Confidence Intervals—Specify the confidence intervals to use for the advanced prediction. See Defining Confidence Intervals.
    • Include Events—Select whether to include events that are associated with the calendar in the prediction. If you select this option, the prediction takes events into account in the data used for training the ML model. See Including Events in Advanced Predictions.
12. Click Save.

    You see a message when the Advanced Prediction is saved.

13. Now you're ready to run the Advanced Predictions you configured. See Running and Scheduling Advanced Predictions.
14. When the job completes successfully, you can download a report with details about the prediction. See Reviewing Detailed Reports for Advanced Predictions.
15. Extend the implementation to include forms, dashboards, navigation flows, and so on, for planners to review and analyze advanced prediction results and to review explainability for a prediction. You can also prepare an Oracle Smart View for Office sheet so planners can review prediction results.

    Explain Prediction, which allows planners to see details about predicted values, is available by default.

    Note that explainability is not available in Smart View.

    Oracle provides a navigation flow for a volume prediction use case. You can use it as an example to customize your implementation for your business use case.

    Tip:

    You can run the advanced prediction using different algorithms, and save each set of results in a different Version. Planners can then select different Versions on a form to see the prediction results from each algorithm.

Advanced Prediction definitions are saved as global artifacts (in the Auto Predict folder under Global Artifacts) and are backed up with the maintenance snapshot.

Additionally, you can export the explainability artifacts of an Advanced Prediction using Migration. From the Home page, click Tools and then click Migration. Artifacts are saved with the Essbase Data in the Plan Type with the prediction.

Videos

Your Goal	Watch This
In this tutorial, you learn which tasks are required before configuring Advanced Predictions. Advanced Predictions takes advantage of machine learning models to generate multivariate predictions to improve your forecasting accuracy.	Setting Up Cloud EPM for Advanced Predictions
In this tutorial, you learn how to configure Advanced Predictions and perform AI-assisted data analysis. Advanced Predictions takes advantage of machine learning models to generate multivariate predictions to improve your forecasting accuracy.	Configuring Advanced Predictions in Cloud EPM

Tutorials

Tutorials provide instructions with sequenced videos and documentation to help you learn a topic.

Your Goal	Learn How
This tutorial supports customers and partners in configuring Advanced Predictions with ML capability to generate more accurate predictions by considering key business drivers. This tutorial shows you how to configure Advanced Predictions and provides implementation recommendations based on business considerations and IPM's solution capabilities.	Configuring and Reviewing Predictions with Advanced Predictions