Dense Relevant Attribute Group Selector

aydin_d

i'm new to rapidminer, and i am trying to get DRAGS from the feature selection extension to work.

unluckily i did not succeed. neither with my own data, nor with sonar from the examples (with the 

class attribute set from special to normal and then converterd from nomial to numeric).

the error is just:

Feb 20, 2018 12:25:06 PM SEVERE: Process failed: null
Feb 20, 2018 12:25:06 PM SEVERE: Here:
Feb 20, 2018 12:25:06 PM SEVERE: Process[1] (Process)
Feb 20, 2018 12:25:06 PM SEVERE: subprocess 'Main Process'
Feb 20, 2018 12:25:06 PM SEVERE: +- Retrieve[1] (Retrieve)
Feb 20, 2018 12:25:06 PM SEVERE: +- Set Data[0] (Set Data)
Feb 20, 2018 12:25:06 PM SEVERE: +- Retrieve (2)[0] (Retrieve)
Feb 20, 2018 12:25:06 PM SEVERE: +- Set Role[1] (Set Role)
Feb 20, 2018 12:25:06 PM SEVERE: +- Nominal to Numerical[1] (Nominal to Numerical)
Feb 20, 2018 12:25:06 PM SEVERE: ==> +- DRAGS[1] (Dense Relevant Attribute Group Selector)

 

can anyone help, e.g. have an working example ?

or is there any way to get at least more infos about where exactly the extension fails?

 

 

aydin_d

i now realize (after using the process the author himself placed somewhere in the internet) that the class

needs labels. thats all. the method (unlike i stupidly thought) is based on class-labels, as the method

will use the average relevance of each feature group detected as criterion. and this relevance is determined

by simple correlation with the class to predict.