X-nominal validation vs svm train and test

side

Merry Christmas to everyone my frineds!

I have a problem using rapid miner. I handle a unbalnced data with 2 classes. I have the arff files and i use them on svm algorithm. When I use X-nominal validation I change the parameters of the algorithm and I have the the results. Although, when I use the svm for traiin and test the prediction label is only one class even though i change the paremeters as in X-validation.

Do you know how to get the same results in svm train and test??

Thanks you!

MariusHelf

Hi,

please post your process setup for test/training set as described in the link in my signature.

How do you create your training/test set? Do you use stratified sampling? Do you get good results with the cross validation?

Best regards,
Marius

side

The problem here is with the arff files so I handled this.

Thank you!