基于贝叶斯网络的P2P网贷借款人信用评价模型
发布时间：2018-08-29 10:21
【摘要】：本文以大数据征信为视角,认为空间维度上借款人多角度、多层次信息的交叉复现,以及时间维度上借款人社会活动信息的持续呈现能够更加准确地反映借款人信用状况,进而构建了基于贝叶斯网络(Bayesian network,BN)的P2P借款人信用评价模型。研究表明,贝叶斯网络为P2P借款人多维信息间的复杂关系提供了统一的表达方式;基于贝叶斯网络推理的样本内信用评价准确率高达87%,提高信用评价的概率值临界点能够显著增强信用评价模型的精准性;样本外信用评价准确率超过90%,增加训练数据集能明显提高信用评价模型的精准性;通过对比不同信息维度模型的评价准确率,也可验证所构建的P2P借款人信用评价模型是稳健的。
[Abstract]:From the perspective of large data, this paper considers that the borrower's multi-angle, multi-level information overlapping and reappearing in spatial dimension, and the continuous presentation of borrower's social activity information in temporal dimension can more accurately reflect the borrower's credit status, and then constructs the P2P borrower's credit evaluation based on Bayesian network (BN). The results show that Bayesian network provides a unified expression for the complex relationship between multi-dimensional information of P2P borrowers; the accuracy of credit evaluation in samples based on Bayesian network reasoning is as high as 87%. Improving the probability critical point of credit evaluation can significantly enhance the accuracy of credit evaluation model; and the accuracy of credit evaluation out of samples can be improved. Increasing the training data set can obviously improve the accuracy of the credit evaluation model. By comparing the evaluation accuracy of different information dimension models, we can also verify that the P2P borrower credit evaluation model is robust.
【作者单位】： 南京大学商学院;
【基金】：国家社科基金重大项目“互联网金融的发展、风险与监管研究(14ZDA043)”的资助
【分类号】：F713.36;F831.2
本文编号：2210904
[Abstract]:From the perspective of large data, this paper considers that the borrower's multi-angle, multi-level information overlapping and reappearing in spatial dimension, and the continuous presentation of borrower's social activity information in temporal dimension can more accurately reflect the borrower's credit status, and then constructs the P2P borrower's credit evaluation based on Bayesian network (BN). The results show that Bayesian network provides a unified expression for the complex relationship between multi-dimensional information of P2P borrowers; the accuracy of credit evaluation in samples based on Bayesian network reasoning is as high as 87%. Improving the probability critical point of credit evaluation can significantly enhance the accuracy of credit evaluation model; and the accuracy of credit evaluation out of samples can be improved. Increasing the training data set can obviously improve the accuracy of the credit evaluation model. By comparing the evaluation accuracy of different information dimension models, we can also verify that the P2P borrower credit evaluation model is robust.
【作者单位】： 南京大学商学院;
【基金】：国家社科基金重大项目“互联网金融的发展、风险与监管研究(14ZDA043)”的资助
【分类号】：F713.36;F831.2
【相似文献】
中国期刊全文数据库 前4条
1 范景学;李晓研;;住房公积金借款人信用风险管理研究[J];辽宁经济;2013年04期
2 毛仙青;信贷登记与贷款卡业务流程[J];浙江金融;2002年01期
3 孙琦;;论我国商业银行信贷风险的控制[J];中国外资;2014年01期
4 ;[J];;年期
中国重要报纸全文数据库 前1条
1 通讯员 陈丽丽邋记者 陈可;外地读大学 家乡获贷款[N];南通日报;2007年
中国硕士学位论文全文数据库 前4条
1 陈丽;p2p借款人信用风险研究[D];华东政法大学;2016年
2 李雪薇;我国P2P网络借贷借款人信用风险影响因素研究[D];内蒙古农业大学;2016年
3 邓皓楠;P2P网贷平台借款人信用风险防范研究[D];湖南大学;2016年
4 陈霞;人人贷借款人信用风险管控研究[D];安徽工业大学;2016年
，本文编号：2210904
本文链接：https://www.wllwen.com/jingjilunwen/huobiyinxinglunwen/2210904.html
