Credit risk evaluation decision modeling through optimized fuzzy classifier

Transparent decision support systems in the finance sector have an important role in the analysis and the decision process. This paper proposes a relatively transparent credit scoring model for evaluating the creditworthiness of the credit applicants through a hybrid data mining approach. The main motivation to apply this approach is to obtain a credit scoring model from a data set that not only have the required performance, but it is relatively interpretable. This objective is achieved through three steps and using complementary soft computing methods. In the first step, the fuzzy system is automatically generated from the data using a fuzzy clustering method. A genetic algorithm is used to increase the performance of the initial fuzzy inference system in the second phase. In the last phase a multi-objective genetic algorithm is applied to achieve two goals: to preserve the accuracy of the fuzzy model to a given value and to enhance the interpretability of the fuzzy model by reducing the fuzzy sets in the rule base. Two datasets from the UCI Machine Learning Repository are selected to evaluate the proposed method.