Comparison of Evaluation Metrics in Classification Applications with Imbalanced Datasets