基于机器学习的信贷欺诈风险预测模型研究  

Research on credit fraud risk prediction model based on machine learning

作　　者：赵志红[1] 杜莹 ZHAO Zhihong;DU Ying(Zhuhai Campus,Beijing Institute of Technology,Zhuhai 519088,China;East China University of Science and Technology,Shanghai 200237,China)

机构地区：[1]北京理工大学珠海学院,广东珠海519088 [2]华东理工大学,上海200237

出　　处：《现代电子技术》2021年第24期177-181,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(11672107);北京理工大学珠海学院科研发展基金(XJZ⁃2019⁃02);广东省普通高校特色创新项目(2019KTSCX217);广东省普通高校青年创新人才项目(2018KQNCX338)。

摘　　要：为提高信贷欺诈风险的预测性能,文中提出一种基于机器学习的信贷欺诈风险预测模型。根据金融系统中存在的信贷欺诈风险类型,计算金融系统信贷数据产生的风险指标。通过确定信贷欺诈风险预测指标权重,设置信贷欺诈风险预测指数。将采集的信贷欺诈数据作为风险预测模型的数据支持,定义信贷欺诈风险产生的信号频谱,并建立风险信息函数,在标准化信贷欺诈风险数据的基础上,计算信贷欺诈风险数据的适应度,完成信贷欺诈风险数据的聚类分析。通过处理信贷欺诈风险数据,采用机器学习算法中的贝叶斯网络节点构建信贷欺诈风险预测模型。最后结合信贷欺诈风险预测流程设计,实现信贷欺诈风险的预测。实验结果表明,通过评价信贷欺诈风险预测的F值指标和AUC值指标,文中设计模型具有更好的预测性能。In order to improve the prediction performance of credit fraud risk,a credit fraud risk prediction model based on machine learning is proposed.According to the types of credit fraud risk in the financial system,the risk indicators generated by the credit data in the financial system are calculated.The credit fraud risk prediction index is set up by determining the weight of the credit fraud risk prediction index.The collected credit fraud data is used as the data support of the risk prediction model,the signal spectrum generated by the credit fraud risk is defined,and the risk information function is established.On the basis of standardized credit fraud risk data,the fitness of credit fraud risk data is calculated,and the clustering analysis of credit fraud risk data is completed.By processing the credit fraud risk data,using the Bayesian network nodes in machine learning algorithm is used to construct the credit fraud risk prediction model.The prediction of the credit fraud risk is realized by combing with the design of credit fraud risk prediction process.The experimental results show that the proposed model has better prediction performance by evaluating F⁃value index and AUC⁃value index of the credit fraud risk prediction.

关 键 词：机器学习 信贷欺诈 风险预测 指数设置 聚类分析 预测模型 金融系统 

分 类 号：TN732-34[电子电信—电路与系统]