Split Validation within Backward Elimination

Muhammed_Fatih_

Hello together, 

I am currently executing a Backward Elimination with contains a simple Naive Bayes by using Split Validation. But the subprocess is iterativeley working by switching between the milestones Naive Bayes --> Apply Model --> Performance and subsequently again starts with Naive Bayes (see attached). 



Is this procesdure normal? 

Thank you in advance for your answers! 

Best regards, 

Fatih

BalazsBarany

Hi Fatih,

in the Validation operators (Split and Cross Validation) you see an output called "ave". This is a bit confusing, but if you move the mouse on it you will see that it expects performance vectors. It could use multiple performances, so for RapidMiner it's a separate type. But you can still use it with the result of the Performance operator.

Regards,

Balázs

BalazsBarany

Hi,

yes, this is a normal procedure. 

You can also use the Split Validation operator with the same operators inside: on the left side your learner (Naive Bayes), and on the right side Apply Model and Performance. This would be equivalent to your solution.

Regards,

Balázs

Muhammed_Fatih_

Hi @BalazsBarany, 

thank you for your answer. I tried this but the outgoing link of Split Validation has no "per" which is necessary for the subprocess of BackwardElimination. 

Do you have a solution for this? 

Best regards, 

Fatih

BalazsBarany

Hi Fatih,

in the Validation operators (Split and Cross Validation) you see an output called "ave". This is a bit confusing, but if you move the mouse on it you will see that it expects performance vectors. It could use multiple performances, so for RapidMiner it's a separate type. But you can still use it with the result of the Performance operator.

Regards,

Balázs