Along with the rapid development of P2P platform,one of the key question is how investors assess the borrowers' credit risk and get effective investment.This paper use the real loan data from a P2P platform,respectively,using logistic regression and random forest to predict the default probability of the loan,then combine the distance measurement model and the kernel regression to evaluate the return and risk of new demand for loans.Due to the amount limit to both the investors and the lenders on P2P platform,this paper also solved a constrained portfolio problem.Data shows that compared with traditional credit rating,the method in this paper improves the loans' yield prediction accuracy and can choose portfolios with higher return.%随着P2P网贷平台的迅速发展,一个关键的问题是投资者如何评估借款人的信用风险并有效投资.本文从网贷平台的真实贷款数据出发,分别用逻辑回归和随机森林预测贷款的违约概率,并通过距离度量模型与核权重相结合评估出新贷款需求的收益和风险.由于P2P网贷平台对放、贷款金额的限制,本文同时解决了一个有约束的投资组合问题.数据结果显示,与传统的信用分级相比,本文的方法提高了贷款收益率预测的准确度,选择出的投资组合收益率更高.
展开▼