Publications

As you know, this is a tough subject.

I'd refer them to a set of papers at the AAAI-2000 workshop on Learning from Imbalanced Datasets. http://borg.cs.dal.ca/~nat/Workshop2000/schedule.html (workshop report at http://www.aaai.org/Press/Reports/Workshops/ws-00-05.html)

Foster's "101" talk/paper is a nice survey of techniques (online at http://www.stern.nyu.edu/~fprovost/home.html#Publications). Before fixing the problem, consider whether overall accuracy really is acceptable.

George