The evaluation of classification models for credit scoring

This paper focused on the problem of the evaluation of classification models for creditscoring. The reason of the model evaluation is to find the optimal solution from variousclassification models generated in an iterated and complex model building process.Classification accuracy is the basic criterion to measure the performance of a model.Classification error rates can be estimated from the test sample. In order to get an accurateestimation, the sample of test data has some requirements: independent from training data,representative for the underlying population, and large enough.Because credit scoring model should be effectively applied to real world problems, othercriterions for evaluating practical scoring models are also important under somecircumstances: the speed, the transparency and the simplicity of the models.Usually credit scoring models classify credit customers into two classes: 'good' and 'bad'. Thetwo types of error rates are...