Improving Classifier Performance Assessment of Credit Scoring Models

In evaluating credit scoring predictive power it is common to use the Re-ceiver Operating Characteristics (ROC) curve, the Area Under the Curve(AUC) and the minimum probability-weighted loss. The main weakness of the rst two assessments is not to take the costs of misclassication errors into account and the last one depends on the number of defaults in the credit portfolio. The main purposes of this paper are to provide a curve, called curve of Misclassication Error Loss (MEL), and a classier performance measure that overcome the above-mentioned drawbacks. We prove that the ROC dominance is equivalent to the MEL dominance. Furthermore, we derive the probability distribution of the proposed predictive power measure and we analyse its performance by Monte Carlo simulations. Finally, we apply the suggested methodologies to empirical data on Italian Small and Medium Enterprisers.