Tenfold Cross-validation Procedure Research Articles

A multicriteria discrimination approach for the credit rating of Asian banks

In this paper we develop a multicriteria decision aid model, to investigate whether it is possible to replicate the credit ratings of Fitch on Asian banks using publicly available data. The model is developed with the Multi-group Hierarchical DIScrimination (MHDIS) approach, following a tenfold cross validation procedure. Five financial variables are selected from a list of nineteen ones through factor analysis. An additional set of five non-financial variables covering ownership, corporate governance, auditing, strength of bank’s franchise and its banking environment is also being used. The results show that equity/customer and short term funding, net interest margin and return on average equity are the most important financial variables. The number of shareholders, the number of subsidiaries and the banking environment of the country in which the banks operate are the most important non-financial ones. In terms of classification accuracies, the results show that the MHDIS model can replicate the credit ratings of Fitch with a satisfactory accuracy and is more efficient than discriminant analysis and ordered logistic regression that are used for comparison purposes.

An ordered logit model for the evaluation of Dutch non-life insurance companies

This paper described the development of a model which determines the financial solidity of Dutch non-life insurers. Since bankruptcies of Dutch insurance companies are very rare, the assessment texts made by the supervisor are used to classify the companies. Because three classes are distinguished instead of two, standard approaches cannot be applied. Instead, an ordered logit model is used. Solvency, profitability, investments, and market share are found to be the most significant aspects. The model correctly classified almost 90% of the companies in a ten-fold cross-validation procedure and 84% in an out of sample test.