An application of a hybrid GA-SVM model to the credit scoring

A robust credit scoring model is badly in demand and is of great significance in the process of constructing a credit system of our country. Since the optimal parameters search of SVM plays a crucial role in building an efficient classification model, we use a genetic algorithm to optimize the parameters of SVM for credit scoring. Experimental results reveal that the GA-SVM model achieved higher accuracy than that of other exiting classifiers, such as statistical methods, back-propagation neural network. The GA-SVM model is proved to be effective in searching the optimal parameters of SVM. The proposed hybrid system has a potential for credit scoring in terms of prediction accuracy and generalization ability.