Hello and Good Day Every One
I'm using Lib-SVM with RBF kernel and when I want to get the probability estimations (I want display probability of each class by changing parameters of command line), the accuracy of the classification drop down with 5 percent when it compares with the accuracy of the classification without the probability estimations. Can any one help me, Why this degradation occur??? I just display result of classifier in details.
Thank you in advance
you can refer to 'LIBSVM: A Library for Support Vector Machines' (Probability Estimates section). the classification algorithm used in probability estimate is a little different from normal svm.
Old versions of WEKA shows that by default, however, you can get probability distribution (only in the case of nominal class attributes) in the latest versions as follows:
In "Classify" panel -> press "More options..." button -> choose "PlainText", then left-click on it, in the popped up dialogue box, select "True" for the "outputDistribution" parameter.
HTH.
Samer
- Badges
- Science topic