基于大数据的电商网贷动态信用评级模型研究——来自“拍拍贷”的经验数据

Research on Big Data-based Dynamic Credit Rating Model for E-commerce Online Lending:Empirical Data from Paipaidai

以“拍拍贷”网贷平台的电商借贷者数据为样本,以违约鉴别能力为准则,采用神经网络判别法和相关分析法对指标进行筛选,构建电商网贷的信用评级指标体系。同时,基于各指标的违约贡献率和AHP方法,采用主观和客观相结合的组合赋权法确定各指标的权重系数。在此基础上,基于时间帧测度电商网贷者的近期信用和长期信用,构建电商网贷动态信用评级模型。该信用评级模型可根据近期信用动态调整长期信用,及时更新用户的信用状况。研究表明,构建的电商网贷信用评级指标体系违约鉴别能力强,历史信息的重要度超过了借款信息、认证信息和个人信息,长期信用将违约样本的信用等级降低,有效降低了信用风险。

Taking the data of e-commerce borrowers of the "Paipaidai" online lending platform as samples,and taking the default discrimination ability as the criterion,this paper screens the indicators,and constructs the credit rating indicator system of e-commerce online lending by using neural network discriminating method and correlation analysis method.Meanwhile,based on the default contribution rate of each indicator and AHP method,the combination weighting method with a combination of subjective and objective factors is adopted to determine the weight coefficient of each indicator.Hence,the time frame is adopted to measure the short term credit and long term credit of E-commercial borrowers,and the dynamic credit rating model of e-commercial borrowers is constructed.The credit rating model can dynamically adjust the long-term credit according to the recent credit and update the credit status of users in time.The research shows that the established E-commerce online credit rating indicator system has a strong ability to identify defaults,and the importance of historical information exceeds that of loan information,certification information and personal information.Long-term credit reduces the credit rating of default samples and effectively reduces the credit risk.