摘要
建立了粗糙集与神经网络集成的贷款风险5级分类评价模型,该模型首先利用自组织映射神经网络离散化财务数据并应用遗传算法约简评价指标;基于最小约简指标提取贷款风险5级分类判别规则以及对BP神经网络进行训练;最后使用粗糙集理论判别与规则库匹配的检验样本风险等级,使用神经网络判别不与规则库任何规则匹配的检验样本风险等级.利用贷款企业数据库698家5级分类样本进行实证研究,结果表明,粗糙集与神经网络集成的判别模型预测准确率达到82.07%,是一种有效的贷款风险5级分类评价工具.
An integrated model of rough set and neural network for five-category classification of loan risk is proposed. The financial data are discretized by using the self-organizing mapping neural network; and the evaluation indices are reduced without information loss through a genetic algorithm. The reduced indices are used to develop the rules for the five-category classification of loan risk, and to train the neural network. The rough set theory is used to determine the category for the test sample which matches all rules in the rule-base. The neural network is applied to separate those test samples which do not match any one rule in the rule-base. 698 loan firms of five-category are selected as test samples. The prediction accuracy of the integrated model combining rough sets and neural network is 82.07%. This verifies the effectiveness of our approach.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2008年第4期759-763,共5页
Control Theory & Applications
基金
国家自然科学基金(70171005)
国家十五攻关项目(2001BA102A06-07-01).
关键词
粗糙集理论
神经网络
贷款风险
分类
rough set theory
neural network
loan risk
classification