A model that you trained using historic data from a source can now be applied to a different set of data. By doing so, a prediction is made for the model field that you selected as the prediction variable. You can execute the prediction process for the data mining methods decision tree, clustering and regression analysis. For clustering, the prediction output would be the best three clusters. The predicted output, in case of a decision tree, is the best three predicted classes and the corresponding probability of the prediction. For the scoring method, regression analysis, the predicted output is the score.
You cannot execute a prediction for association analysis because training is already the result of this data mining method. You can export the association rules determined in training to SAP CRM (see the section Association Analysis).
· The model must be trained
· To execute the prediction, the analysis process must be activated
To create an analysis process for prediction:
1. Choose Create
2. Select an application from the drop-down menu and select Okay. Your analysis process will be assigned to the appropriate folder on the left side of the screen
3. Enter the description for the analysis process
4. Drag a data source into the work area and enter the detail settings in the dialog box that appears
5. Drag the relevant prediction icon, that is, source for transformation, in the work area
6. Connect the two nodes with the mouse
7. Select the data mining method and choose Properties from the context menu. Alternatively, you can double click on data mining node to make the settings in the dialog box that appears
For more details, see Transformations for an Analysis Process.
The prediction result is determined by the system and then displayed in graphic form.
You can load the determined values into the master data in SAP BW (see Loading Mining Results into SAP BW)