A Multi-objective Credit Risk Assessment System Based on Support Vector Machines Ensemble Model

Description

Credit risk assessment is receiving renewed interest from both academics and the business community. Efficient estimations of credit risk can greatly reduce losses. Most existing models focus only on improving classification accuracy and have performed better on identifying good customers than bad customers. In practice, however, the cost of misclassification of a bad customer as good is generally more than that of a good customer classified as bad. So the credit granting strategies for financial institutions are based on multiple objectives such as minimizing loss while maintaining a competitive market share.The purpose of this research is to construct a multi-objective credit risk assessment system with a support vector machines (SVM) ensemble model. The SVM ensemble model will be compared with other traditional models in terms of different criteria. To improve the practical use of the research results, real world datasets from banks will be used to verify the performance of the system and interpret the decisions to form decision rules, which will be provided to risk mangers in financial institutions for inspecting and learning.