| 926 | 23 | 194 |
| 下载次数 | 被引频次 | 阅读次数 |
近年来,我国消费金融发展迅速,但同时也面临着更加复杂的欺诈和信用风险,为了更好地对消费金融中借贷客户的信用风险进行监测,本文提出了基于稀疏结构连续比率模型的风控方法。相对于传统的二分类模型,该模型的特点是可以处理借贷客户被分为三类或三类以上的有序数据,估计系数的同时能从众多纷繁复杂的数据中自动筛选重要变量,并在变量筛选过程中考虑不同子模型系数的结构特征。通过蒙特卡洛模拟发现,本文所提出的稀疏结构连续比率模型在分类泛化误差和变量筛选上的表现都较好。最后将本文提出的模型应用到实际的消费金融信用风险分析中,针对传统征信信息不足的借款人,通过引入高频电商消费行为数据,利用本文提出的高维有序多分类模型能有效识别借款人的信用风险,可以弥补传统征信方法的不足。
Abstract：Consumer finance has been developing rapidly in recent years in China,but it is also faced with risks of more complex fraud and credit. In order to better monitor the credit risk of consumers finance debtors,this paper proposes a new risk control method based on sparse structure continuation ratio model. Compared with traditional binary classification model,this model can handle ordinal response with three or more than three categories,which can estimate coefficients and meanwhile automatically conduct variable selection taking into account the structure information of coefficients in different sub-models. The Monte Carlo simulation results suggest that the proposed model has a good performance on variable selection and classification prediction.Finally,the proposed model is applied to the real consumer finance credit risk analysis,and we find that for debtors with insufficient credit information from traditional sources,by bringing high-frequency e-commerce behavior data and using the proposed high-dimensional ordinal multi-classification model,we can effectively identify the credit risks of the debtors and make up for the shortcoming of traditional credit method.
[1]方匡南，范新妍，马双鸽．基于网络结构Logistic模型的企业信用风险预警[J]．统计研究，2016,33(4):50-55.
[2]方匡南，赵梦峦．基于多源数据融合的个人信用评分研究[J]．统计研究，2018,35(12):92-101.
[3]方匡南，章贵军，张惠颖．基于Lasso-logistic模型的个人信用风险预警方法[J]．数量经济技术经济研究，2014(2):125-136.
[4]胡心瀚，叶五一，缪柏其．上市公司信用风险分析模型中的变量选择[J]．数理统计与管理，2012,31(6):1117-1124.
[5]石庆焱，靳云汇．多种个人信用评分模型在中国应用的比较研究[J]．统计研究，2004,21(6):43-5.
[6]王小燕，方匡南，谢邦昌．Logisitic回归的双层变量选择研究[J]．统计研究，2014,31(9):107-112.
[7]王正位，周从意，廖理，等．消费行为在个人信用风险识别中的信息含量研究[J]．经济研究，2020,55(01):149-163.
[8]叶强，刘作义，孟庆峰，等．互联网金融的国家战略需求和关键科学问题[J]．中国科学基金，2016(2):150-158.
[9]余乐安，张有德．基于关联规则赋权特征选择集成的信用分类研究[J]．系统工程理论与实践．2020,40(2):366-372.
[10]Archer K J,Williams A A. L1Penalized Continuation Ratio Models for Ordinal Response Prediction Using High-dimensional Datasets[J]. Statistics in Medicine,2012,31(14):1464-1474.
[11]Choi H.,Koo J,Park C. Fused Least Absolute Shrinkage and Selection Operator for Credit Scoring[J]. Journal of Statistical Computation and Simulation,2015,85(11):2135-2147.
[12]Churchill G A,Nevin J R,Watson R R. The Role of Credit Scoring in the Loan Decision[J]. Credit World,1977,3(3):6-10.
[13]Cox D R. Regression Models and Life-tables[J]. Journal of the Royal Statistical Society:Series B(Methodological),1972,34(2):187-202.
[14]Liu J,Ma S,Huang J. Integrative Analysis of Cancer Diagnosis Studies with Composite Penalization[J]. Scandinavian Journal of Statistics,2014,41(1):87-103.
[15]Mc Cullagh P. Regression Models for Ordinal Data[J]. Journal of the Royal Statistical Society:Series B(Methodological),1980,42(2):109-127.
[16]Myers J H,Forgy E W. The Development of Numerical Credit Evaluation Systems[J]. Journal of the American Statistical Association,1963,58(303):799-806.
[17]West D. Neutral Network Credit Scoring Models[J]. Computers&Operations Research,2000(27):1131-1152.
[18]Wiginton J C. A Note on the Comparison of Logit and Discriminant Models of Consumer Credit Behavior[J]. Journal of Financial&Quantitative Analysis,1980,15(3):757-770.
[19]Zhang C. Nearly Unbiased Variable Selection Under Minimax Concave Penalty[J]. The Annals of Statistics,2010,38(2):894-942.
[20]Zou H,Li R. One-step Sparse Estimates in Nonconcave Penalized Likelihood Models[J]. The Annals of Statistics,2008,36(4):1509-1533.
基本信息:
DOI：10.19343/j.cnki.11-1302/c.2020.11.005
中图分类号:F832.4;F224
引用信息:
[1]张晶,张喆,方匡南,等.基于稀疏结构连续比率模型的消费金融风控研究[J].统计研究,2020,37(11):57-67.DOI:10.19343/j.cnki.11-1302/c.2020.11.005.
基金信息:
教育部人文社科研究青年基金“基于半监督学习的消费金融风控方法与应用研究”（20YJC910004）;; 全国统计科学研究重大项目“多源数据融合的无监督学习方法及其应用”（2019LD02）;; 国家自然科学基金面上项目“基于多源信息融合的高维分类方法及其在信用评分中的应用”（72071169）
2020-11-25
2020-11-25
2020-11-25