遗传算法优化神经网络及信用评价研究

研究关于公司神经网络信用评估问题的现状,提出遗传算法辅助网络训练策略(优化后的网络称为进化网络),克服传统网络建模中产生的局部极小缺陷。建立了适合于我国商业企业的信用评分指标体系;然后依据该指标体系建立了基于进化神经网络的信用评估模型;最后,利用样本公司实际指标数据对该模型的评分效果进行了比较研究。

A Strategy of Optimizing Neural Networks by Genetic Algorithm and Its Application on Credit Scoring

Studying the international actuality of corporation credit scoring,a strategy of optimizing neural networks by genetic algorithm is proposed,thus the deficiency of local minimizing appeared while training the networks is handled.The optimized networks is named evolutive network.Then,after some distinguished indices selected and sample designed carefully,by use of the evolutive neural network,credit-score models applied to Chinese business corporations are built up in this study.Finally,comparison research of the credit-score models’ scoring powers are carried out by use of actural index data.