Keeping attribute selection after optimal selection — Is it possible?

ar4o

Hi again!

 

I have 2 small questions this time. I have created a model with Optimize Selection with Cross-Validation inside and it shows good results. 

 

1) Does it overfitt the model? 

2) Can I and if yes then how can I apply this model to another dataset? Without optimize selection everything was clear but with it...

 

Working with SVM for text classification.

 

Thank you in advance!

Thomas_Ott

Overfit can happen, I would toggle on the "leave one out" parameter and test again.

 

W.R.T to the Optimize Selection, you should be able to write out the weights to a repository and use a Select By Weights operator in your scoring process and load them back in.