Learning from imbalanced data sets with boosting and data generation