Classification Performance Assessment For Imbalanced Multiclass Data

The article discusses the importance of evaluating diagnostic systems, particularly in the context of multiclass datasets with imbalanced class distributions. It introduces the Imbalanced Multiclass Classification Performance (IMCP) curve as a measure of performance for these systems, and demonstrates its effectiveness in assessing individual class performance in medical diagnosis. Empirical experiments with real-world data show that the IMCP curve and area under the curve are reliable indicators of classification quality.