Orange Forum • View topic - Linear SVM weight for feature ranking

Linear SVM weight for feature ranking

A place to ask questions about methods in Orange and how they are used and other general support.

Linear SVM weight for feature ranking

Postby bgbg » Thu Oct 08, 2009 12:20

Hello,
in this paper http://jmlr.csail.mit.edu/proceedings/papers/v3/chang08a/chang08a.pdf the maintainers of libsvm describe how they successfully used linear SVM to rank features. I know that orange SVM implementation uses libsvm. However, I couldn't find a way to obtain these weights. Am I missing anything?
Code: Select all
data = orange.ExampleTable('someFile.tab')
learner = orngSVM.SVMLearner(kernel_type=orngSVM.SVMLearner.LINEAR)
classifier = learner(data)
weights = classifier. ?????????

Postby Ales » Tue Oct 13, 2009 10:15

Try this:

Code: Select all
weights = orngSVM.getLinearSVMWeights(classifier)


Return to Questions & Support