Running the Analysis
- To run a Radial Basis Function analysis, from
the menus choose:
Figure 1. Radial Basis Function: Variables tab 
- Select Customer category [custcat] as the dependent variable.
- Select Marital status [marital], Level of education [ed], Retired [retire], and Gender [gender] as factors.
- Select Age in years [age] through Number of people in household [reside] as covariates.
- Select Adjusted Normalized as the method for rescaling covariates.
- Click
the Partitions tab.
Figure 2. Radial Basis Function: Partitions tab 
By specifying relative numbers of cases, it's easy to create fractional partitions that would be difficult to specify percentages for. Say you want to assign 2/3 of the dataset to the training sample and 2/3 of the remaining cases to testing.
- Type 6 as the relative number for the training sample.
- Type 2 as the relative number for the testing sample.
- Type 1 as the relative number
for the holdout sample.
A total of 9 relative cases have been specified. 6/9 = 2/3, or about 66.67%, are assigned to the training sample; 2/9, or about 22.22%, are assigned to testing; 1/9, or about 11.11% are assigned to the holdout sample.
- Click the Output tab.
Figure 3. Radial Basis Function: Output tab 
- Deselect Diagram in the Network Structure group.
- Select ROC curve, Cumulative gains chart, Lift chart, and Predicted by observed chart in the Network Performance group.
- Click the Save tab.
Figure 4. Radial Basis Function: Save tab 
- Select Save predicted value or category for each dependent variable and Save predicted pseudo-probability for each dependent variable.
- Click OK.