P2P网贷平台借款人信用风险评估

Assessment of Borrowers’ Credit Risk in Online P2P Lending

为提高P2P网贷平台对借款人信用风险评估的准确性,利用主成分分析模型,用8个主成分反映18个指标,依据熵权法和方差百分比赋予权重,得到借款人是否违约还款的临界值,依据该值对借款人是否违约还款进行评估;再利用Logistic回归模型,得到影响借款人违约发生率的4个指标,其按影响程度从大到小的排序为借贷子等级、借贷目的、过去两年里超过30 d以上未及时还款的次数、贷款收入比。对11 383条交易数据的识别结果表明,在阈值为50%时,评估模型对正常还款的拟合率为99.6%,对违约还款的拟合率为71.3%,整体的拟合准确率为96.04%。

A principal component analysis was used to improve the accuracy of credit risk assessment of borrowers by P2 P online lending platforms.18 indicators were reflected by eight principal components, and weights assigned according to entropy weight and percentage variance to obtain the critical value by which to evaluate whether the borrower defaults on repayment.Four indicators influencing the incidence of borrower default were then obtained using logistic regression,which were ranked in descending order of influence as sub-grade of loan, purpose of loan, times of overdue repayment over 30 days in the past two years, and loan income ratio.The results from identification of 11,383 transactions show that at 50% threshold, the fitting rate of the evaluation model is 99.6% for normal repayment, 71.3% for default repayment, and the overall fitting accuracy is 96.04%.