Classifying data with a support vector machine in RapidMiner

supportvectormachine

Hello,

Coming from IBM SPSS Modeler I'm experiencing some troubles modeling the same approach in RapidMiner.

I basically followed this tutorial on the IBM SPSS Modeler documentation:

https:// www. ibm. com/support/knowledgecenter/en/SS3RA7_18.2.1/modeler_tutorial_ddita/clementine/example_svm_intro.html

in which a dataset of specific variables from cells gets taken into an SVM in order to identify the class of a certain cell type.

I've attached the CSV as a file but do not know how to create an SVM in RapidMiner that does use all of the attributes except Class and Patient ID in order to determine the Class.

Can anyone hint me in the right direction?

Thanks in advance.

lionelderkrikor

Hi @supportvectormachine,

You can find a working process using a SVM model and your dataset  with : 
 - the attribute ID set as "id"
 - the attribute Class set as "label"

The model is inside a Cross Validation operator in order to estimate the performance of your model.

Hope this helps,

Regards,

Lionel

lionelderkrikor

Hi @supportvectormachine,

The Optimize Parameters (Grid) operator has been made for you !!!
Take a look at the process in attached file and you can play with the parameters of this operator.

The results look like that : 


Hope this helps,

Regards,

Lionel

Marco_Barradas

Hi @supportvectormachine please take a look at these videos from our academy

https://academy.rapidminer.com/learn/video/svm-intro
https://academy.rapidminer.com/learn/video/svm-demo
https://academy.rapidminer.com/learn/article/understanding-support-vector-machines

Hope you find them useful.

lionelderkrikor

Hi @supportvectormachine,

You can find a working process using a SVM model and your dataset  with : 
 - the attribute ID set as "id"
 - the attribute Class set as "label"

The model is inside a Cross Validation operator in order to estimate the performance of your model.

Hope this helps,

Regards,

Lionel

supportvectormachine

Hello @lionelderkrikor

Thank you very much for your answer. I already had a similar solution, but was stuck at the following problem: My SVM had an error where it would not accept any polynomial attributes.

What did you in order for your SVM to accept polynomial attributes?

lionelderkrikor

@supportvectormachine,

You're welcome ! 
Your attribute called "BareNuc" is initially of type "Polynominal". I used "Guess Types" operator to transform
it from "Polynominal" type to "Integer" type which is supported by the SVM model : 



Regards,

Lionel
 

supportvectormachine

Thank you very much @lionelderkrikor

Additionaly, say if I wanted to try out several different kernels in my SVM and compare their performance, how would I go about this?

My model looks like this as of now.

lionelderkrikor

Hi @supportvectormachine,

The Optimize Parameters (Grid) operator has been made for you !!!
Take a look at the process in attached file and you can play with the parameters of this operator.

The results look like that : 


Hope this helps,

Regards,

Lionel

supportvectormachine

Thank you so much! That helped a lot.