BP算法在信用风险分析中的应用.pdf

BP 算法在信用风险分析中的应用 李行风 1 张博群 2 徐建东 1 1. 暨南大学数学系，广东，广州，510632 2. 华南理工大学交通学院，广东，广 州，510640 摘要：按照企业的财务状况、经营状况以及过往的信用记录 分为“信用好”和“信用差”两个小组。对于每一家贷款企业， 主要考虑能反映该企业的还款能力、盈利能力、经营效率和资本 结构等 7 个财务比率做为分析变量。 关键词：BP 算法；信用风险评价模型；信用风险分析 Application on BP Algorithm in Credit Risk Analysis Li Xingfeng1 Zhang Boqun2 Xu Jiandong2 1). Department of Mathematics, Jinan University, Guangzhou, Guangdong, 510632, China 2). College of Traffic and Communications, South China 得到国家自然科学基金（705682 ）和广东省自然科学基金（31906 ）资助 １/ 10 University of Technology ， Guangzhou, Guangdong, 510640, China Abstract：The research establishes a credit-risk evaluation model based on back-propagation algorithm. The model has been applied to evaluate the credits of 80 applicants in a commercial bank of our country in 2001. The 80 data are separated into two groups: a “good credit” group and a “badcredit” group according to their finance, management and previous credit records. To each applicant, it mainly is considered seven financial rates that can reflect its debt paying ability, profitability, quality of management and capital structure. The BP network is trained 100 times, 390 times and 800 times respectively. The simulations show that, when the network is trained 800 times, it enters steady state and the performance function reaches optimal value. And the classification accuracy rate is 98.75%. In addition, the paper also presents a learning algorithm and steps of the BP network. ２/ 10 Keywords：BP algorithm；Credit risk evaluation model； Credit risk analysis 1. 引言 信用风险分析是指信用评级机构根据科学的指标体系，采用 严谨的分析方法，运用简明的文字符号，对被评级单位履行经济 责任的能力及其可信任程度进行客观公正的评价，并确定其信用 等级的一种经济活动。对贷款企业进行信用评级，建立相应的信 用风险评价模型，并运用此模型预测某种