Indentifying and eliminating insignificant attributes

Repletion

Hello!
Within statistics we usually operate with a 5% significance value for determining which attributes are significant or insignificant in the model. Thereby one can remove insignificant attributes and get a "simpler" model. How would one go ahead doing this in Rapidminer and is there an automated process for it?

lionelderkrikor

Hi @Repletion,

I'm not aware of an implementation of the feature your describe in RapidMiner.
But more generally, what you describe - get a "simpler" model - is called feature selection.

In RapidMiner you can apply feature selection by : 
 - Enabling Automatic Feature Selection and choosing your option (simple/balanced/accurate) in Auto-Model.
 - Using directly in your process the Automatic Feature Engineering operator.
 - Using the feature weights operator(s) (Weight by Information Gain,  Weight by Correlation etc.) associated with the Select by Weights operator.

Hope this helps,

Regards,

Lionel