基于RF的Elastic Net-Logistic个人信用违约风险评估  

Assessment of Elastic Net-Logistic individualcredit default risk based on random forest

作　　者：陈倩 贺兴时[1] 杨新社[2] CHEN Qian;HE Xingshi;YANG Xinshe(School of Science,Xi’an Polytechnic University,Xi’an 710048,China;School of Science and Technology,Middlesex University,London NW44BT,UK)

机构地区：[1]西安工程大学理学院,陕西西安710048 [2]密德萨斯大学科学与技术学院,英国伦敦NM44BT

出　　处：《西安工程大学学报》2021年第3期116-122,共7页Journal of Xi’an Polytechnic University

基　　金：国家自然科学基金(12001417);陕西省科技厅软科学项目(2019KRM072);陕西省教育厅自然科学专项(19JK0373)。

摘　　要：以南德信贷数据为基础,针对信贷数据中解释变量维数高、类型丰富、好坏客户数量不均衡等特点,通过分析影响个人信用违约风险的因素,构建了一种基于随机森林(random forest,RF)的Elastic Net-Logistic个人信用违约风险评估模型。并与传统Elastic Net-Logistic模型、基于RF的Lasso-Logistic模型进行比较,证明所提出的模型表现更优。为更进一步验证模型有效性,采用澳大利亚信贷数据进行实例分析。结果表明,该模型的违约召回率与分类精度比Elastic Net-Logistic模型和基于RF的Lasso-Logistic模型分别提高了0.88%、8.78%和0.79%、6.06%。验证了基于RF的Elastic Net-Logistic模型对信贷数据有更好的分类效果。Based on the credit data of South Germany,in terms of the characteristics of the high dimensionality of the explanatory variables,the rich types,and the unbalanced number of good and bad customers in the credit data,the factors that affect personal credit default risk were analyzed,the Elastic Net-Logistic individual credit default risk assessment model based on random forest(RF)was constructed and compared with the traditional Elastic Net-Logistic model and the Lasso-Logistic model based on RF.The results prove that the proposed model presented is better.In order to further verify the validity of the model,the Australian credit data were used.The results show that the default recall rate and classification accuracy of the model are improved by 0.88%,8.78%and 0.79%,6.06%respectively compared with the Elastic Net-Logistic model and RF-based Lasso Logistic model respectively,which verifies that the model has better classification effect on credit data.

关 键 词：个人信用风险 随机森林 Elastic Net-Logistic模型 Lasso-Logistic模型 

分 类 号：F832.479[经济管理—金融学] F224