Polynomial to binomial special attribute

Thiru

dear all,  Im trying a case to compare ROCs.  Compare ROCs does not accept polynomial special attribute and i thought
of using only two labels , i.e. binomial.  can I do this without importing modified data again?  

can anyone help?  

  ( In fact,  I was looking for an operator like 'select example' ( something like select attribute), so that I can select examples I want. seems it is  not available.   )

thank you. 

IngoRM

  ( In fact,  I was looking for an operator like 'select example' ( something like select attribute), so that I can select examples I want. seems it is  not available.   )

Did you try "Filter Examples" or "Filter Example Range" for this?
Cheers,
Ingo

Telcontar120

You can also create a new label where you map multiple values of your existing polynominal to only two new values and then predict that instead.

Thiru

IngoRM  thanks your reply.  Using " filter example " - we can select the particular rows.  But the Rapidminer still considers the classification as polynomial and not as binomial even though it considers the rows we want for the Model classifcation.

Thiru

Telcontar120  , thanks for your reply.  Tried to generate new attribute.  The existing special attribute name: ' Fault '     which has  three values  'no', 'yes', 'bypass'.  Now I want to generate new attribute same as 'fault" but  without having the value "bypass"  , so that  the new attribute can become binomial.  Tried to use text information + matches  function.  It is not working.  could you pl help. thanks. 

varunm1

Hello @Thiru

When you tried using "Filter Examples", you are still seeing bypass (0) with zero values in label (Fault) statistics? If this is the issue, can you add "Guess Types" operator after the filter examples and select option "Include special attributes" and then add "Nominal to Binominal" operator selecting "include special attribute" option and select the "Fault" attribute.

I attached a sample process related to the Iris data set that has three label values, I am filtering out one value and converted the label column from Nominal to Binominal. You can import this process by going to FILE --> Import process.

Do let us know if this works.

Telcontar120

You can also try adding the "Remove Unused Values" operator after you filter out the examples you don't want to include (the ones with "bypass").  Sometimes RapidMiner doesn't update the metadata after a filter, so the Guess Types by itself may still register it as a polynominal rather than a binominal.  After the Remove Unused Attributes there should only be two values left which will allow it to bet set as a binominal.