【作者】 祝煜； 梁雪春； 肖迪；

【Author】 ZhuYu Liang Xuechun XiaoDi College of Automation,Nanjing University of Technology,Nanjing,210009,China

【机构】 南京工业大学自动化学院；

【摘要】 本文提出了基于动态聚类算法的RBF神经网络信用风险预测模型,利用动态聚类算法确定RBF网络的隐含层节点,不仅聚类速度快,而且隐含层节点数的优化提高了网络的利用效率。定义了广义重要度欧氏距离用于算法中的距离计算,使得算法具有全局优化的聚类结果,根据穆迪、安德尔违约概率曲线定义了信用评级风险系数等指标。最后,以南京某商业银行提供的数据为依据,利用matlab为工具平台,建立基于动态聚类的RBF神经网络模型。实证分析表明:基于动态聚类的RBF神经网络信贷预测模型对违约小企业的判别准确率较高,可为银行有效地甄别高风险企业。从而说明此预测模型的设计是准确可靠的。更多还原

【Abstract】 In this paper,the credit risk forecasting model based on dynamic clustering algorithm RBF neural network is presented.This clustering algorithm has the characteristic of speedy learning. Optimization of the number of hidden layer node improves the efficiency of network.The general important degree Euclidean distance algorithm for calculating the distance is defined;this makes the algorithm optimize clustering results.According to Moody,Alder’s default probability curve,we define the indices such as the risk coefficient of credit rating.At last,we carry on the empirical analysis on the base of the data which come from some commercial bank in Nanjing in China,and use matlab tool platform then found the small enterprises credit forecasting based on dynamic clustering algorithm RBF neural network. The result shows that the model based on dynamic clustering algorithm RBF neural network can make the accuracy rate higher;and the banks can effectively avoid high-risk enterprises.The simulation indicates that the model is credible and accurate.更多还原