0

I perform a classification of two labels using libsvm. But I don't get good results for the default parameters of SVM kernel type = linear. Can any one please tell me a way to find best parameters for SVM linear kernel type

  • There is no standard way to do this. Maybe you need to do more data preprocessing. In general, you do not provide enough information about your data, so nobody is really able to answer your question. – knb Oct 20 '15 at 07:25
  • I am using features extracted from Facebook :) –  Oct 21 '15 at 07:40
  • Grid search is a common way. I would recommend to use the RBF kernel. But stillt it has more parameter you need to vary.. – Jürgen K. Aug 26 '16 at 10:50

1 Answers1

1

The linear kernel depends on the C parameter.

You could perform a grid-search for this parameter in order to find the 'best' matching one for your given dataset.

For weka the procedure is described here.

rzo1
  • 5,561
  • 3
  • 25
  • 64