基于非均衡模糊近似支持向量机的P2P网贷借款人信用风险评估及应用

现实中P2P网贷平台可信用户和违约用户的样本分布具有非均衡性,且投资者对分类错误持有不同接受程度.本文通过使用双边权重误差测量方法和映射距离选择正负样本误差项的隶属度,构建了基于非均衡模糊近似支持向量机(DFPSVM)的P2P网贷借款人信用风险评估模型.然后,提出了借款人信用评分及评级方法.最后,借助人人贷平台借款人信用信息进行了实证分析,结果表明所构建的模型与其他模型相比具有更好的适应能力和较高的分类准确度,能有效减少样本非均衡性对分类结果的影响,显著增加负类样本分类的准确率.获得的人人贷平台借款人的信用得分、信用等级及违约率分布能够为平台控制违约风险及投资者决策提供帮助.

The borrowers' credit risk assessment in P2P platform based on fuzzy proximal support vector machine and its application

It is the fact that the sample distributions of trusted users and default users in P2P platform are unbalanced and the investors have different acceptable degrees in classification error. This paper establishes the disequilibrium fuzzy proximal support vector machine (DFPSVM) to assess the borrowers' credit risk in P2P platform by using the bilateral weighted error measuring method and mapping distance to measure fuzzy memberships of the positive and negative samples error term. Next, it proposes the borrowers' credit scoring and credit rating model in P2P platform. Finally, the empirical results based on the borrowers' information in Renrendai platform show that the proposed DFPSVM model has better generalization ability and higher classification accuracy than other existing models. It can effectively reduce the effect of disequilibrium samples and increase the classification accuracy of negative samples. The obtained borrowers' credit score, credit rating and the distribution of default rate is helpful to control the default risk of P2P platform and support the decision making process.