Personal Credit Evaluation Combined Model Based On Decision Tree And Neural Network

China is currently in a transition period of social market economic system, its market economy has been rapidly developed these years, however the construction of social credit system is not compatible with the development of the economic growth. Personal credit rating plays an important role in social credit system. The accuracy and efficiency of credit scoring model has a close bearing on the profits of financial institutes and the prosperity of financial market. Due to the imperfections of the personal credit system, laws and regulations, as well as the management deficiencies, the development of China’s personal credit received serious constraints. Therefore, establishing a comprehensive personal credit rating system has become an important task of all the financial institutions.As one of the most widely used classification algorithms of data mining, decision tree algorithm is characterized by simplicity, high efficiency and clear structure. However, as one of the acknowledged remarkable credit rating method which owned high prediction accuracy, neural network is criticized for its unknown structure and instability.In this paper, we acquire the real personal credit data samples of a German commercial bank from the internet, use the decision tree algorithm based on C5.0, through data collection, marking the attribute index, the selection of training samples, and then establish the raw personal credit evaluation decision tree model. Then, through adjusting the pruning severity, using misclassification costs and boosting, we form the optimal decision tree. The factors of great importance which were extracted from the optimal decision tree would be input the neural network model, then the combined model output the final classification results. From the model’s prediction results and the comparison with the single BP neural network, the combined model has improved stability and accuracy by removing the interference of some unimportant factors. And the method to choose attributes with decision tree, which is based on the principle of maximizing the information entropy gain ratio, truly enhance the model’s interpretability. This combined model has certain guiding function in the practice of the lending decisions to the commercial bank credit officers, it could provide support for credit.