Test model for different outcome variable

aksaha

Hi,
 I have rapidminer deep learning predictive model for a output variable, I want to see how to the model performs to a different outcome variable using the same learning information. What would be the best operator to use ?
Thanks

sgenzer

hi @aksaha I'm not sure exactly what you mean here. Do you want to rebuild the model with a new predicted class, or do you want to score the model on new data with the same predicted class? The latter is easy - just use Apply Model. The former is not hard either - just use Set Role to set a new predicted class and build the model again.

Does that help?

Scott

varunm1

Hello @aksaha

If you are looking to predict multiple label columns, which I think you are trying to do. You need to use a "Multi-label modeling" operator. You can see the tutorial in that operator "help" window. 

Note: Multi-label modeling creates multiple models based on the number of outcome variables (attributes) you are trying to predict.

 I want to see how to the model performs to a different outcome variable using the same learning information

Generally, a model trained for one outcome variable will support prediction for that variable. If you are looking for "transfer learning" then I am not sure if rapidminer supports that. 

If you are trying to predict the same label column for new data, then @sgenzer suggestion works