Variable Selection Method

Use this option to specify the method for selecting the final predictor variables.

This parameter can be specified either using radio buttons or a drop-down menu, depending on the predictive model chosen.

Possible choices are described below:

Variable Selection Method 1	Definition
None	Choose this to use all of the variables as predictors. Caution : Choosing this option can result in long run times when there are thousands of variables.
Stepwise	Choose this option to select predictor variables using a forward stepwise method.
Penalized	Choose this option to apply a penalty proportional to the sum of squares of the coefficients.
Genetic Algorithm	This method selects a fixed subset size by considering and refining multiple random candidate sets. Genetic Algorithm is not available in Interactive Mode and tends to be slower than Stepwise . Caution : Genetic Algorithm processing is computationally intensive.
Forward	This method performs a forward selection.
Backward	This method performs a backwards elimination.
Score	This method performs a best subset search based on Score statistics with respect to the specified number of predictors included in the model. When searching for the best subset, all of the possible combinations are screened and might take a long time to finish. This option does not work with a class predictor.

---

1

Note: The methods available to you vary among different predictive modeling processes. Not all are available for every process.

To Specify the Variable Selection Method:

Select the desired option either by clicking within the appropriate radio button or using the drop-down menu.