Borderline Oversampling in Feature Space for Learning Algorithms in Imbalanced Data Environments