Approaches to classification of imbalanced data - a case study on automotive materials