Creating or rebuilding a model

Add an SPSS® Model process box to an HCL® Campaign flowchart to create a new modeling stream or use an existing modeling stream to rebuild or refresh a model. You can then view and configure the model build options in IBM® SPSS Modeler Advantage Enterprise Marketing Management Edition. These settings are used when the model is built during an HCL Campaign automated process flow.

Before you begin

If you are using an existing modeling stream, it must be a compatible .str file that can be loaded into IBM SPSS Modeler Advantage Enterprise Marketing Management Edition. The prerequisites for using a modeling stream that is created in IBM SPSS Modeler Advantage Enterprise Marketing Management Edition are described in the "Sharing Streams" topic in the HCL SPSS Modeler Advantage Guide. In addition, the modeling stream must contain a numeric key field named SPSS_AudID.

About this task

Complete the following steps to configure a model process.

Procedure

  1. In HCL Campaign, open a flowchart in Edit mode.
  2. Drag process boxes from the palette to the flowchart workspace. Connect them as you normally would to select a set of audience members.
    For example, use the Select, Segment, and Merge process boxes.
  3. Drag the SPSS Model process box from the palette to your flowchart workspace.
  4. Connect a process to the SPSS Model process box.
    The output of the process that is upstream of the SPSS Model process provides input to the SPSS Model process.
    If you connect a Select process box to the SPSS Model process box, the Select process provides the set of audience members to use for building the model.
  5. Double-click the SPSS Model process box in the flowchart workspace.

    The SPSS Model Process Configuration dialog opens.

  6. Select the input data source, which is the input cell.
  7. Select whether to create a modeling stream or to rebuild a model in an existing modeling stream.
    OptionDescription
    Create new modeling stream Create a modeling stream.
    Note: The stream name must be unique. By default, the modeling stream name is based on a concatenation of the campaign name, the campaign code, the flowchart name, and the SPSS Model process name. You can change the name to any unique value, but it is useful to retain a traceable link to the process box that created this modeling stream. If a stream with that name exists, a warning is displayed.
    Use an existing modeling stream Click Browse to select a modeling stream. In the Open Model File window, the default is to choose a modeling stream from the IBM SPSS Collaboration and Deployment Services repository.

    Click Upload Local File to select a modeling stream from your local hard drive.

  8. Depending on the model type, select what you are trying to predict in the Target field.
    • For a Predictive model, use the Target field.
    • For an Association model, use the Content field.
    • For a Clustering model, use the Clustering Evaluation field.
  9. Specify the variables to be considered for modeling.
    • When you build a model, you can select any of the Candidate fields, which are based on the table mappings that are available in the flowchart.
    • When you use an existing stream, the Stream Field column is automatically populated. The Match, Up, and Down buttons are not displayed if the Stream Fields are not populated.
  10. Optional: To change the default model build options, click Open Modeler Advantage Session.

    IBM SPSS Modeler Advantage Enterprise Marketing Management Edition opens in a new window. The Modeling tab is displayed. The modeling stream might be partially or completely configured and ready to build a model, but a model has not been built yet.

    The SPSS Model process box is now configured to use an existing modeling stream, which is the stream that was just created. HCL Campaign passes the default target variable and the audience ID and key, and creates a dummy input data file to build the model.

    1. Select the modeling type.

      Make sure the default Predictive model type is selected.

    2. Optional: On the Modeling tab, select the values for this model process.
      • Model-specific settings: such as the percentage of data to use for the training set versus the test set
      • Build settings: enabling or disabling automatic data clean up
      Note: Make sure the numeric key field SPSS_AudID is included.
    3. Optional: Click the Data tab and view the data configuration.

      If you are creating a modeling stream from the SPSS Model process box, the Data tab is configured with data from the flowchart. If you are using an existing modeling stream, the Data tab is configured with data from the modeling stream. You can change data sources and use any features, such as creating new variables. Any data source changes are in effect only for the current session; they are overwritten on any run of the SPSS Model process.

    4. Optional: Click Build Model after you configure the settings.

      You can build the model only if data is available on the SPSS Modeler server, such as if the SPSS Model process box was previously run.

    5. Save your changes and exit IBM SPSS Modeler Advantage Enterprise Marketing Management Edition to return to the SPSS Model Process Configuration dialog.

      The modeling stream was saved in the IBM SPSS Collaboration and Deployment Services repository.

  11. Click OK to save and close the SPSS Model Process Configuration dialog.

    HCL Campaign verifies whether the settings are valid and a model is ready to be built.

    • If the settings are valid, the SPSS Model process box displays the status configured.
    • If the settings are not valid, the SPSS Model process box displays the status unconfigured with an appropriate error message.