PERSONAL CREDIT SCORING MODEL RESEARCHBASED ON THE RF-GA-SVM MODEL

Abstract

The importance measure of variables in the random forests algorithm is used to carry outa rank ordering to the importance of variables, so as to extract feature attributes on this basis. The feature attributes are regarded as inputs to conduct parameter optimization in order to support vector machine (SVM) model by using the genetic algorithm, building the classifier model by selecting the parameter with the highest accuracy of 5-fold cross-validation. The classifier model is utilized for empirical research, and the results show that the classifier is better than random forest classifier and support vector machine classifier in its higher classification accuracy.