基于神经网络及Logistic回归的混合信用卡评分模型

Credit Evaluation Model of Credit Card by Using the Hybrid Model of Neural Network and Logistic Regression

如何更好地对信用卡申请人进行识别和判断, 提高银行预防和抵抗信用卡风险的能力, 是所有银行迫切需要解决的问题。为了做出更高效、智能的判断, 为决策者提供有效的决策支持, 从而提高银行信用卡审批过程中资信评估的正确率及效率。本文提出结合神经网络及 Logistic 回归技术的混合模型, 结合神经网络预测精度高及 Logistic 回归稳定性高的特性,对信用卡进行信用评分。实证分析的结果表明,混合模型的预测精度确实比单独使用神经网络模型和 Logistic 回归的预测精度高, 证明了该方法确实是有效可行的。

It is urgent for banks to solve problems as to how to identify and judge applicants of？credit card, and improve capability of preventing and resisting risk of credit card. To provide effective decision support to policymaker to judge the credit card efficiently and intellectually will increase accuracy and efficiency of credit evaluation in the process of credit card approval. This paper evaluates credit card by using the hybrid model of Neural networks and Logistic regression with their characters of high accuracy of neural networks and high stability of Logistic regression. Real case shows that the accuracy of hybrid model is higher than using neural networks and logistic regression separately. Therefore, this method is feasible and effective.