Prediction Score SVM

simbolonsimbolon Member Posts: 1 Newbie
Hello,

Im new on rapidminer;
Currently, i am not able to 
get the prediction value when using LIBSVM.
I tried to understand that the calculation is 
f(x)=(w,x)+b.

Could you please explain how can I get the prediction value columns?
Thank you in advance.

Answers

  • jacobcybulskijacobcybulski Member, University Professor Posts: 391 Unicorn
    I am not sure if I understand your difficulty. However, from the prediction viewpoint, SVM is like any other binomial classifier (LIBSVM can be used as estimator as well). So you need to specify a label attribute, possibly use a cross-validation (or hold out if wished to), use LIBSVM as part of training, and Apply Model and Binomial Classification Performance as part of testing. Then run.

    Note however that you've picked one of the most difficult models to work with. I suggest to start with a decision tree or logistic regression to start with, get the process right and only then move to SVMs. As SVM uses a kernel (you will need to specify which kernel type you want to work with), it will need to be optimised for this kernel type, and this means playing with C and other kernel specific parameters.
Sign In or Register to comment.