Strategy ONE

Importing predictive models for MDX cubes

MicroStrategy allows you to integrate data from MDX cube sources such as SAP BW and Microsoft Analysis Services. Once integrated into MicroStrategy as MDX cubes, you can use the MicroStrategy reporting and analysis features to further analyze your MDX cube source data. For information on integrating MDX cube sources into MicroStrategy, see Importing MDX cubes in the MDX Cube Reporting Help.

This also includes performing predictive analysis on your MDX cube data. The steps below show you how to import a predictive model into your MDX cube.

Prerequisites

Before the Import Data Mining Model option is available:

  • You must have the following user privileges: Use Metric Editor (Developer) and Create Application Objects (Common Privileges)

  • You must be logged into a project.

  • You have imported an MDX cube, as shown in the MDX Cube Reporting Help, to act as the dataset for the predictive model.

  • You have created and trained a predictive model from the dataset, using MicroStrategy or a third-party application.

To import the model

  1. In MicroStrategy Developer, log in to a project that is connected to an MDX cube source.

  2. If you are using read only mode for the project, from the Schema menu, clear the Read Only Mode option to switch to edit mode.

    Only one user can be editing a project at a given time. Therefore, if someone else is modifying the project, you cannot use the MDX Cube Catalog.

  3. From the Schema menu, select MDX Cube Catalog.

    • If you have a single MDX cube source database instance created for the project, the MDX Cube Catalog opens.

    • If you have multiple MDX cube source database instances created for the project, a Database Instance dialog box opens. Select a valid MDX cube source database instance, click OK, and the MDX Cube Catalog opens.

  4. Click the Cube Mapping tab.

  5. From the Catalog\Cube drop-down list, select the MDX cube to import a predictive model to.

  6. From the Edit menu, select Import Data Mining Model. The Import Data Mining Model dialog box opens.

  7. Click ... (the Browse button) to locate the Data Mining Model to import. Once you select the file, the fields in the table at the bottom of the dialog box are automatically populated.

  8. In the Predictor Type column, select or clear the type of predictive metrics to be created. By default, all outputs defined by the model you chose above are selected. If outputs are not defined in the selected PMML model, then MicroStrategy offers the predicted value (or score) and probability (or confidence) outputs where applicable.

  9. In the Name column, change the default names provided for each predictive metric if you want.

  10. In the Aggregation Function column, specify the desired aggregation functions for the selected predictive metric types, if you want to change from the default functions.

  11. Click OK.

  12. Select an existing or create a new destination folder for the predictive metrics. All predictive metrics will be created in this folder.

  13. Click OK.

  14. The model is imported, which may take a few minutes depending on the size of the model and the number of predictive metrics selected. All predictive metrics are created within the Compound Metrics of the MDX cube.

    If a predictive metric already exists in the MDX cube with the same name, you are prompted whether to overwrite it. If you choose not to, you are returned to the Import Data Mining Model dialog box to rename the predictive metric.

Importing the model into MicroStrategy results in the creation of a predictive metric. This metric contains the imported PMML model, and an expression that invokes the appropriate MicroStrategy function to execute the model. The next step is to use the predictive metric in reports.

Related Topics