基于遗传算法的BP神经网络在企业资信评估中的应用  

Application in creditrating evaluation based on BP neural network

作　　者：孟凡超[1] 张洪伟[1] 徐剑[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《计算机应用研究》2007年第8期301-303,305,共4页Application Research of Computers

摘　　要：提出了一种新的企业资信评估方法。通过把神经网络和遗传算法有机地结合起来,既克服了传统BP网络训练时间长、易陷入局部极值的缺点,又利用遗传算法提高了网络全局收敛的效率。该模型采用C#.NET+SQL server 2000实现。实验结果表明,基于遗传算法的BP神经网络系统对企业资信评估有着良好的性能。This paper presented a new credit rating evaluation algorithm by combining genetic algorithm and back propagation （BP） neural network, It not only conquered the demerits such as long time net-training and easily running into local optimum in traditional BP network, but also improved the convergent efficiency of network by using GA. The model was implemented with C#, NET and SQL server 2000, The rationality of experiment results indicate that BP-based neural network can evaluate corporations accurately and efficiently by credit rating.

关 键 词：遗传算法 反向传播神经网络 资信等级评估 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]