WEKA has profound support for kNN classifiers (many different distances and etc.)
Unfortunately WEKA doesn't support multi-label problems.
One of the solutions can be to use binary relevance approach.
I am not sure whether it's a correct workaround? What do you think?