To configure a Model process

Configure a Model process to generate response models with the IBM® Campaign modeling engine. The Model process generates a runtime model file (.rtm). You can use the .rtm file with the Score process to determine which customers or prospects are most likely to respond.

Open a flowchart in Edit mode.

Drag the Model process from the palette to your flowchart.

Configure and connect the input to the Model process to identify responders and non-responders.

For example, you can use a Select process to select all contacts for an offer, then use a Segment process to identify responders and non-responders.

As another example, you can use one Select process to identify responders. To produce a list of non-responders, use a second Select process to identify people who were contacted, then use Merge to exclude responders. You then connect the Merge process (non-responders) and the Select process (responders) as input to the Model process.

Double-click the Model process in the flowchart workspace.

The process configuration dialog opens.

Use the Source tab to select the fields to use for modeling.

Select responder and non-responder cells from the Responder and Non-Responder lists. The lists are populated with input cells that are connected to the Model process.

All input cells must have the same audience level.

For Variables to be Used for Modeling, select the fields to be used by the Model process during model generation. You can click Use All and allow the Model process decide which set of inputs are most effective for modeling. However, you can speed up the modeling process by eliminating variables that do not add value to the model. For example, you can eliminate fields that contain the same value for all records or different values for all records.

If you are in doubt as to the predictive value of a variable, include it and allow the Model process decide whether it should be used.

Use the Method tab to specify how to generate the model.

Select one of the following options to determine how the data mining algorithms should arrive at the best model.

Best Model In: (Default) Specify a time limit for modeling. The Model process retains the best models built in the time period that you specify. The default is three hours.

Best Model Among: Specify the number of candidate models to build. The Model process will retain the best of these models. The default is 20.

Algorithm to Use: Select the type of modeling to perform. You can select one, several, or all algorithms. Selecting All Algorithms (the default) results in a more accurate model, but it might take longer.

Max # of Models to Keep: Specify the maximum number of models to keep. The default is 5. The more models that you keep, the more disk space is required. Each of the top models is saved by appending a number sign (#) to the end of the model file name. If more than one model is kept, an index is appended to the base file name, indicating the rank of the model.

Model File Name: Specify the absolute path name for the runtime model file (.rtm) that will be created, or click Browse to navigate to a file. When you run the Model process, it generates a NAME.rtm model file, which you can use for scoring with the Score process.

(Optional) Use the General tab to assign a name and descriptive note.

Click OK.

The process is now configured. You can test run the process to verify that it returns the results you expect.