A hybrid model for credit risk assessment: empirical validation by real credit data
This article examines which hybridization strategy is most appropriate for assessing credit risk in the dynamic financial world. As such, we use large new datasets and develop different hybrid models by combining traditional and modern artificial intelligence statistical methods based on cluster feature classification and selection approaches. We find that a multilayer perceptron (MLP) combined with discriminant analysis or logistic regression (LR) can dramatically improve classification accuracy compared to other simple and hybrid classifiers. In particular, the results of our empirical analysis, of our test of statistical significance and of our test of expected cost of misclassification confirm the superiority of the hybrid LR + MLP classifier based on clustering to improve the precision of the prediction in the maximum performance criteria. To verify the effectiveness and viability of the proposed model, we use three unbalanced data sets: Chinese farmer credit, Chinese small and medium-sized enterprise (SME) credit, and German credit. We also use Australian credit data for further authentication and robustness check. The first two datasets are private and large in size, while the second two are mostly used, publicly accessible and small in size. Thus, our results are relevant for many areas of credit risk, such as modeling the credit risk of SMEs, farmers and consumers.