Prerequisites for building and scoring models

Make sure that the prerequisites are met before you build and score models in IBM® Campaign.

In IBM® SPSS® Modeler Advantage Marketing Edition, modeling configuration and metadata are stored in proprietary stream files, which are called modeling streams.

When you configure a new SPSS Model process box in an IBM® Campaign flowchart, you can create a new modeling stream or use an existing modeling stream.

For the SPSS Score process, you can use existing modeling streams that were created in IBM® SPSS® Modeler Advantage Marketing Edition.

Modeling streams must meet the following requirements:

The modeling stream includes SPSS_AudID as a numeric key field. This field is created automatically for any modeling stream that is created through an SPSS Model process box. If you are manually creating a modeling stream, you need to ensure that your data source contains a numeric SPSS_AudID field that uniquely identifies each individual in the data set. For scoring, the SPSS_AudID key field must be selected as scoring output.

The modeling stream uses a single data source node.

The deployment metadata indicates which branch is the Model Refresh branch.

The selected modeling node is either an automated modeling node or a Decision List node.

A Type node is present to define the target field.

You can download modeling streams by accessing the IBM® SPSS® Collaboration and Deployment Services repository. The download process is described in the IBM® SPSS® Collaboration and Deployment Services User's Guide.

Data fields

IBM® Campaign passes user data fields to IBM® SPSS® Modeler Advantage Marketing Edition through .csv data files. These data files are written to a shared folder for modeling and scoring.

The data can contain numeric, text, and date fields. However, IBM® SPSS® Modeler Advantage Marketing Edition understands date fields in numeric form only. If your modeling data includes dates, you must convert them to numeric values before that data is passed to IBM® SPSS® Modeler Advantage Marketing Edition. You can use the system-defined Date () macro in IBM® Campaign to convert dates into Julian format.

Dummy data files

If you create a modeling stream in the SPSS Model process box, IBM® Campaign generates dummy data according to the types for the candidate fields that are specified in the SPSS Model process box. The dummy data file is stored in the common shared folder and is used to create a stream. The purpose of the dummy file is to pass field names and data types to IBM® SPSS® Modeler Advantage Marketing Edition so that IBM® SPSS® Modeler Advantage Marketing Edition can understand the field types in the modeling stream.

If you open the modeling stream in IBM® SPSS® Modeler Advantage Marketing Edition before you run the SPSS Model or SPSS Score process box, only dummy data is available for the SPSS Score process box. You must run the processes that provide the input cell and run the SPSS Model process to see real data in any of the IBM® SPSS® Modeler Advantage Marketing Edition data overviews or previews.

The dummy file cannot be used to build the modeling stream. Actual data for modeling is passed only when the SPSS Model process box is run in the IBM® Campaign flowchart.

Test runs

Test runs do not require that predecessor process boxes have run successfully, but the runs might not generate meaningful results. For test mode, only partial user data is provided to test flowchart function. Other than that, a test run of the SPSS Model or SPSS Score process box is the same as a production run. Both types of runs rebuild the model that is based on the latest data from IBM® Campaign.