Abstract
Although various algorithms have widely been studied for bankruptcy and credit risk prediction, conclusions regarding the best performing method are divergent when using different performance assessment metrics. As a solution to this problem, the present paper suggests the employment of two well-known multiple-criteria decision-making (MCDM) techniques by integrating their preference scores, which can constitute a valuable tool for decision-makers and analysts to choose the prediction model(s) more properly. Thus, selection of the most suitable algorithm will be designed as an MCDM problem that consists of a finite number of performance metrics (criteria) and a finite number of classifiers (alternatives). An experimental study will be performed to provide a more comprehensive assessment regarding the behavior of ten classifiers over credit data evaluated with seven different measures, whereas the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) and Preference Ranking Organization METHod for Enrichment of Evaluations (PROMETHEE) techniques will be applied to rank the classifiers. The results demonstrate that evaluating the performance with a unique measure may lead to wrong conclusions, while the MCDM methods may give rise to a more consistent analysis. Furthermore, the use of MCDM methods allows the analysts to weight the significance of each performance metric based on the intrinsic characteristics of a given credit granting decision problem.
Highlights
The 2007–2008 global financial crisis and the recommendations on banking regulations have attracted the growing interest of institutions in credit and operational risk management, which has become a key determinant of success because incorrect decisions may lead to heavy losses
Even a more obvious example is for the results over the Thomas database: the Bayesian belief network, logistic regression, multilayer perceptron (MLP), and support vector machine (SVM) achieved the highest rates when using the accuracy, the naïve Bayes classifier was the model with the highest true-negative rate and geometric mean, and MLP and random forest were the best algorithms on the F-measure
The present analysis supports the synergetic application of multiple-criteria decision-making (MCDM) techniques for the performance assessment of credit granting decision systems
Summary
The 2007–2008 global financial crisis and the recommendations on banking regulations have attracted the growing interest of institutions in credit and operational risk management, which has become a key determinant of success because incorrect decisions may lead to heavy losses. Credit granting decision can be expressed in the form of a two-class prediction problem in which a new case has to be assigned to one of the predetermined classes according to a set of input or explanatory attributes. These attributes or variables gather a diversity of information that summarizes both socio-demographic features and financial status of the credit applicants, whereas the classifier gives an output based on their financial solvency.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
