基于Levenberg-Marquardt神经网络的企业资信评估方法研究  

Corporation Credit Rating with Neural Network Based on Levenberg-Marquardt Algorithm

作　　者：蔡秋茹[1] 柳益君[1] 蒋红芬[1] 罗烨[1] 叶飞跃[1] 

机构地区：[1]江苏技术师范学院计算机科学与工程学院,常州213001

出　　处：《计算机与数字工程》2009年第7期154-156,180,共4页Computer & Digital Engineering

基　　金：常州市"831工程"资助项目(编号:KYZ06002);江苏技术师范学院基金重点资助项目(编号:KYY06036)资助

摘　　要：企业资信评估问题是一个复杂的非线性问题,而神经网络技术可实现非线性关系的隐式表达。文章提出将基于Levenberg-Marquardt算法的多层前馈型神经网络用于资信评估,并通过MATLAB软件及其神经网络工具对其进行仿真计算。实验结果表明,企业资信神经网络评估模型收敛速度快,准确率较高,具有一定的实用价值。Corporation credit rating is a complex nonlinear problem. It has become a hot topic in the application of neural networks to solute the problem of corporation credit rating. In this study, the neural network based on Levenberg-Marquardt algorithm is applied to the problem. Matlab software and its neural network toolbox are used to make network compute. The experiment results show that the neural network model is effective and efficient.

关 键 词：神经网络 LEVENBERG-MARQUARDT算法 资信评估 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]