Credit classification of legal customers of an Iranian bank with KFP-MCOC

Credit risk is one of the most important banking risks that is due to not paying principal and interest of loans. Measuring credit risk is important; because not measuring it lead to increasing volume of doubtful accounts and unexpected future losses. In this research a model was proposed that based on linear and nonlinear optimization. This model is finding a separating hyperplane which classify 85 good and bad borrower customers of Iranian’s bank. This customers are all in Tehran Stock Exchange (TSE). In order to improving the model we used kernel functions, data fuzzification and penalty factors in it. The results show that the best model among linear and nonlinear models with linear, polynomial, sigmoid and RBF kernels, is a linear optimization model with sigmoid kernel function that has accuracy of 80% and recall of 100%.