Feature Selection Options

 

Information-based measure

Calculates the strength of the relationship between each input field and the target (last) field. The strength of the relationship is measured using information theory. This measure is only applicable when the target field is categorical or discrete numeric.

 

Correlation coefficients

Calculates the strength of the relationship between each input field and the target (last) field. The strength of the relationship is measured using correlation coefficients. This measure is only applicable when the target field is numeric.

 

Display feature selection results

A report dialog can be displayed to show the scores for each field.

 

Select best fields for output

Selects the best fields for use in a later part of the current stream.

 

Number of fields to select

Enter the number of best-scoring fields that will be selected. Alternatively, click on the spin control to increase or decrease the current value. The default value is 1.

 

Continuous field step size

The information gain measure can be slow for continuous numeric fields with a large number of records. Increasing the step size will increase the speed of the algorithm but may reduce the accuracy of the results.

You can click on the spin control to increase or decrease the current step size for the information-based method. The default value is 1. We recommend that you leave the setting at 1 unless the process is very slow for some fields, in which case increase the value until you obtain an acceptable rate of progress.