基于QUBO模型的信用评分卡组合优化研究  

作　　者：许宏飞 牟玲星 许文杰 邓姗姗 唐亚林 

机构地区：[1]湖南科技学院,湖南永州425199 [2]湘潭大学,湖南湘潭411100

出　　处：《科技创新与应用》2024年第19期43-46,共4页Technology Innovation and Application

基　　金：湖南省教育厅项目(21C0701);2023年永州市指导性科技计划项目(2023YZ002);湖南科技学院科学研究项目(23XKYZZ05);湖南科技学院科学研究项目(20XKY060)。

摘　　要：银行贷款审核制度对于保障银行财产安全、防范信贷风险和维护金融市场秩序具有重要作用。为提高审核效率和准确性,该文引入二维0-1变量,构建选择1张信用评分卡和3张信用评分卡的银行最大收益模型。借助变量线性组合的方程将不等式约束转变为等量约束,通过换元法将二维变量表示为一维变量,利用0-1变量的性质添加惩罚项将目标函数转化为无约束的QUBO模型。应用模拟退火算法进行快速求解并为银行信用贷款资格审核工作提供指导。Bank loan audit system plays an important role in ensuring bank property safety,preventing credit risk and maintaining financial market order.In order to improve the efficiency and accuracy of auditing,this paper introduces a two-dimensional 0-1 variable to construct a bank maximum return model which selects one credit scorecard and three credit scorecards.With the help of the equation of linear combination of variables,the inequality constraints are transformed into equivalent constraints,the two-dimensional variables are expressed as one-dimensional variables by the substitution method,and the objective function is transformed into an unconstrained QUBO model by adding a penalty term according to the properties of 0-1 variables.The simulated annealing algorithm is applied to solve the problem quickly and provide guidance for the bank credit loan qualification verification.

关 键 词：信用评分卡 最大收益模型 0-1二值变量 QUBO模型 银行信用卡 

分 类 号：O29[理学—应用数学]