基于粗糙集与遗传算法集成的企业短期贷款违约判别  

Short-term loan default prediction based on integration of rough sets and genetic algorithm

作　　者：柯孔林[1] 冯宗宪[2] 

机构地区：[1]浙江工商大学金融学院,杭州310035 [2]西安交通大学金禾经济研究中心,西安710049

出　　处：《系统工程理论与实践》2008年第4期27-34,共8页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(70171005)

摘　　要：建立了粗糙集和遗传算法集成的企业贷款违约判别模型.该模型首先利用FUSINTER方法离散化财务数据,并应用遗传算法约简评价指标,进而基于最小约简指标提取违约判别规则,最后对企业短期贷款检验样本进行违约判别.利用贷款企业数据库558家样本企业进行交叉验证技术的实证研究,结果表明,与多元判别分析l、ogistic、BP神经网络等违约判别模型相比,粗糙集和遗传算法集成的违约判别模型是一种更为有效和实用的信用风险评估工具.Integrated model of rough sets and genetic algorithm for short-term loan default prediction is proposed in this paper. Firstly, financial data is discretised by using FUSINTER method and evaluation variables is reduced through genetic algorithm. And then, this reduced information is used to develop default classification rules. Lastly, a set of rules are used to discriminate between healthy and default testing samples. Financial data of 558 loan firms is selected, the effectiveness of integrated model of rough sets and genetic algorithm was verified by cross-validation experiments comparing with multiple discriminant analysis, logistic analysis and BP neural network approach.

关 键 词：粗糙集理论 遗传算法 短期贷款 违约判别 

分 类 号：F830.51[经济管理—金融学]