The competition uses a partial dataset of PAKDD2009 competition. You are required to predict a bad customer using the given set of features. There are 40K records for training and 10K for testing.