我国P2P网络借贷借款人信用风险的识别研究

发布时间：2018-01-21 19:20

本文关键词： P2P 风险识别多项Logistic回归模型 BP神经网络模型　出处：《吉林财经大学》2017年硕士论文　论文类型：学位论文

【摘要】：自2007年P2P网络借贷传入中国以来,经历了短暂的时间便获得了惊人的发展。P2P网络借贷借助于互联网经济的快速发展,运营平台的数量呈现出井喷式增长。2007年至2015年之间,P2P网络借贷行业运营平台从开始的第一家到现在的第2595家,并且2015年单年增加了1020家,绝对的增加量远远超过过去任何一个时期。据统计,平台平均每个月的成交额为492.6亿元,并且一直处于扩张的状态。但是,由于P2P网络借贷平台具有进入门槛低、缺乏行业标准、监管不严格等特点,致使P2P网络借贷平台存在严重的信息不对称问题,从而P2P平台会出现我们经常听到的“跑路”等现象。据统计,截止到2015年6月,国内出现信用问题的P2P网络借贷平台达到550家之多。目前,国外的金融市场给P2P网络借贷行业提供了一个健康的发展环境,此外,全面的监管机制也促使了国外P2P网络借贷行业的发展。但是P2P网络借贷公司的发展受到实际的一定限制,由于P2P属于新兴起的行业,风险控制能力和传统的银行是无法相比的。一般而言,网络借贷主要风险有以下两种:基本风险和特定风险。基本风险有法律风险、信用风险和监管风险等;特定风险则包括投资风险、信息不对称风险等。而信用风险危害最为严重,很多借贷平台还是采用的传统商业银行的信用识别模型,并没有根据网络借贷的特点开发出更适合网络借贷的信用风险识别模型。本文基于“拍拍贷”的数据,结合国内外现有的信用风险识别方法,建立了信用风险识别的指标体系。在SPSS和MATLAB操作软件的基础上,应用多项Logistic回归方法,以魔镜等级“E”为参照等级,分别建立A-F等级的多项回归方程,得到每个等级的显著性影响因素。后续建立了神经网络模型,首先将变量进行归一化处理,然后确定模型中的阈值和权值,通过不断的改变权值,达到了对模型训练及仿真的目的。最后对两种方法进行了对比和总结。根据本次研究,以上方法都可以对借款人信用风险进行有效的识别,根据计算结果可以有效的识别借款人的信用风险等级,并且验证了两个模型对网络借贷借款人信用风险的识别能力。后续的研究中,对如何提升借款人信用风险的识别能力提出了建议。为加强网络借贷行业的风险识别能力,应丰富多层次的信息认证指标,并加强政府机构对P2P网络借贷行业的有效监管。
[Abstract]:Since 2007, P2P lending into China, experienced a short period of time will get rapid development of.P2P lending network alarming with the help of the Internet economy, between the number of operating platform showing a growth spurt in.2007 years to 2015, P2P network lending industry operating platform to the present 2595th from the beginning of the first, and 2015 single year increase of 1020, the absolute increase in the amount of far more than ever. According to statistics, the average monthly turnover platform for 49 billion 260 million yuan, and has been in the expansion of the state. However, due to the P2P network lending platform has low barriers to entry, the lack of industry standards, not strict supervision characteristics. The P2P network lending platform serious asymmetric information problems, so that the P2P platform will appear we often hear the "run away" phenomenon. According to statistics, by the end of June 2015, China The P2P network lending platform credit problems reach as much as 550. At present, the international financial market to provide a healthy environment for development, in addition to P2P network lending industry, a comprehensive regulatory mechanism also contributed to the development of foreign P2P network lending industry. But limited the development of P2P network lending company by the actual. Because P2P belongs to a new industry, risk control ability and the traditional bank cannot be compared. In general, the main risk of lending to the network has the following two types: basic and particular risks. The basic risk and legal risk, credit risk and supervision risk; the specific risks include investment risk, the risk of information asymmetry. The credit risk is the most serious harm, many credit recognition model of lending platform or the use of the traditional commercial bank, and not according to the characteristics of the development of network lending is more suitable for the network by The credit risk identification model. Based on the loan pat loan data, combined with the credit risk existing recognition methods at home and abroad, set up the index system of credit risk identification. Based on SPSS and MATLAB operating software, using multinomial Logistic regression method, such as "E" in the mirror as the reference level, a number respectively. The establishment of A-F level regression equation, are significant factors for each grade. The neural network model was set up, the variables are normalized, and then determine the threshold and weight in the model, by changing the weights continuously, reached the training and simulation model. Finally, the two methods were compared and summary. According to this study, the above method can effectively recognize the credit risk of the borrower, according to the calculation results can effectively identify the borrower's credit risk level, and The two models have been verified the recognition ability of the network lending the borrower's credit risk. The follow-up study, some suggestions on how to improve the credit risk of the borrower's ability to identify risk identification is proposed. To strengthen the network lending industry, we should enrich the multi-level index authentication information, and strengthen the government's effective supervision of P2P network lending industry.

【学位授予单位】：吉林财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.6;F832.4

【参考文献】