https://github.com/scikit-learn/scikit-learn/issues/4587 Is this still on the cards? I'm willing to work on this. I was thinking on lines of creating a synthetic dataset, using blobs, moon etc like the clustering example. For the clustering, I was thinking of using DBSCAN. For the classifier (naive bayes/random forest) apart from the clusters obtained, one of the target classes could be 'noise' (points not clustered by DBSCAN) Let me know if this sounds right. Chirag NagpalUniversity of Punewww.chiragnagpal.com
------------------------------------------------------------------------------
_______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
