0
$\begingroup$

I ran SVM with several kernels on my data. RBS has the best performing results. The task is similar to the text classification. I wonder how I can explain why RBS is actually the best kernel for my task. I can simplify the task a bit, and compare RBS with only linear. I user libsvm to train and test.

$\endgroup$
1
$\begingroup$

Interesting, since it is recommended NOT to use the RBF kernel for text classificiation, see for example here.

When you compared the models, did you use cross validation using train, test and validation set? See here why this is needed. If you need to choose the `best' model, you need an additional test set, and to optimize the parameter for the RBF kernel and the parameter C, you need the validation set ;).

| cite | improve this answer | |
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.