I am a Ph'd student from Algiers and I am very happy to join your
community. I study in image processing and i want to use weka in this
domain. I am very recent in using weka so i have many questions to ask
In my project i have two steps to do:
In the first step is to use a multiclass (SVM) classifier in order to
classify a set of images by getting a training model; and in the
second step is: from the test outputs (like the svm margins or
decision functions f(xi)) for each instance, to measure the quality of
So my aim problem with weka is i have no access to these margins in
order to get the quality measure.
So my question is: is there any way to get these margins (or decision
margins) at the outputs of svm classifier.
I will be very grateful if anyone can help me. Best regards.
Wekalist mailing list
Send posts to: [hidden email] To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
All you can get in WEKA is class probability estimates. Turn on calibration using logistic models in SMO, which will fit logistic regression models to the outputs of the SVMs and perform pairwise coupling to yield multi-class probability estimates.
The distributionForInstance(Instance) method will give you the class probability estimates for a given test instance.