个人信用评估的Logistic-RBF组合模型  

Personal credit scoring based on Logistic and RBF combined model

作　　者：姜明辉[1] 谢行恒[2] 王树林[1] 温潇[3] 

机构地区：[1]哈尔滨工业大学管理学院,哈尔滨150001 [2]宁波工程学院,浙江宁波315016 [3]清华大学人文社会科学院国际问题研究所,北京100084

出　　处：《哈尔滨工业大学学报》2007年第7期1128-1130,共3页Journal of Harbin Institute of Technology

基　　金：哈尔滨工业大学技术.政策.管理(TPM)国家哲学社科创新基地资助项目(HTCSR06T06)

摘　　要：针对个人信用评估中单一模型存在的不足,提出了利用组合预测模型进行个人信用评估的方法.基于不同单一模型在个人信用评估中所体现的优势,选择具有代表性的Logistic回归和径向基函数神经网络方法,建立了2种单一评估模型,在此基础上构建了基于二者的组合模型.利用某商业银行的数据进行2类模式的分类,应用结果表明,组合模型有效地提高了预测的精确性和模型的稳健性,对于商业银行控制消费信贷风险具有更好的适用性.Aiming at the insufficiencies of single models in personal credit scoring, this paper presents a method for personal credit scoring by using combining forecast. Based on the advantages of single method, this paper chose typical Logistic regression and RBF neural network to construct two single models and then constructed a combining forecast model. Using the constructed models to classify the consumer credit data from one commercial bank, the application result indicates that the combining forecast model increases the accuracy effectively as well as model＇ s stability which presents more applicable for commercial banks to keep away from consumer credit risks.

关 键 词：LOGISTIC回归 神经网络 组合预测 个人信用评估 

分 类 号：F224.0[经济管理—国民经济]