大数据背景下网络借贷的信用风险评估——以人人贷为例

在大数据时代,网贷平台每天流动着海量交易数据。为充分利用这些数据控制信用风险,运用数据挖掘算法建立了信用风险评估模型。由于网贷数据多为非平衡数据,所以通过多次尝试使用SMOTE算法进行处理,提高了模型评估性能。研究发现:随机森林模型更适合用于信用风险评估,其次是CART、ANN、C4.5。用户的婚姻、房/车产(贷)等信息重要程度较低,而公司规模、工作时间等信息,历史借款、信用评分等信用档案信息在信用风险评估中尤为重要。

The Evaluation of the Borrower's Credit Risk in Peer-to-Peer Lending under the Background of Big Data:Evidence from RenRen Dai

Massive transaction data is flowing on the Peer‐to‐Peer lending platforms every day in the age of big data .For the purpose of making the most of these data to control the credit risk effectively ,we established the credit risk evaluation model of Peer‐to‐Peer lending using data mining methods .Moreover , due to the imbalance of the data , we decided to use the synthetic minority over‐sampling technique (SMOTE) to improve the performance of the credit risk model after several tries .The empirical study found that Random Forests is more suitable for the evaluation of credit risk .CART ,ANN and C4 .5 also perform well .In addition ,the borrower's marital status and possession of house ,car ,mortgage and auto loan is of no importance ,but their personal information (company size ,employment length ,etc .) and credit information (loan information ,credit score ,etc .) play an important role in the evaluation of credit risk .