P2P网贷中借款人信用风险评估

发布时间：2018-04-26 01:14

本文选题：P2P网络借贷 + 借款人信用评价体系　；参考：《北方工业大学》2017年硕士论文

【摘要】：在信息化时代的今天,"互联网+"模式引领了整个产业结构,互联网金融发展的更是风生水起。作为经济转型的重要方式,P2P网络借贷的到来无疑是锦上添花。P2P网络借贷是基于网络平台,实现个人与个人之间的资金往来、借贷交易,通过将小额度资金聚拢,而获取更大收益的一种商业模式。作为新兴行业,虽然国家不断颁布相关法律条例,严格监管P2P网络借贷平台的运作模式,但仍然存在着诸多问题,最为重要的是关于借款人的信用风险评价。针对此问题的研究无论是对P2P网络借贷平台本身,还是对于投资方而言都有着十分重要的实用价值与指导意义。本文系统地论述了我国P2P网络借贷的主要运作模式、特点以及相关风险,以P2P借款人信用风险为核心,以网络借贷平台抓取的数据为蓝本,定性与定量方法双管齐下,科学的筛选影响借款人信用风险的主要因素,确定适当的自变量,从而构建借款人信用风险评估体系,并进行描述性的统计分析。进一步分别引用决策树与Radial-Basis Function径向基函数(RBF)神经网络两种分类算法,建立P2P网络借贷借款人信用风险评估模型,进行仿真训练,有效地对借款人信用进行量化评估。最后针对两种算法在指标选取、预测精度两个方面的输出结果进行对比后发现:利率、还款期限、借款总额以及还清笔数这四种变量是两种算法共同重要的评价指标;选取较少评价指标的决策树模型的预测精度要高于综合考虑了所有评价指标的RBF神经网络模型。然而,单独识别信用良好的借款人群,RBF神经网络模型的优势更大。
[Abstract]:In the information age, the "Internet" mode has led the entire industrial structure, and the development of Internet finance is booming. As an important way of economic transformation, there is no doubt that the arrival of P2P network lending is an added bonus. P2P network lending is based on a network platform to realize the exchange of funds between individuals, loan transactions, and by gathering small amounts of funds. And a business model that makes more money. As a new industry, although the state has constantly promulgated relevant laws and regulations to strictly supervise the operation mode of P2P network lending platform, there are still many problems, the most important is the credit risk evaluation of borrowers. The research on this problem has very important practical value and guiding significance for P2P network lending platform and investors. This paper systematically discusses the main operation mode, characteristics and related risks of P2P network lending in China. Taking the credit risk of P2P borrowers as the core and the data captured by the network lending platform as the blueprint, the qualitative and quantitative methods are combined. Scientific screening of the main factors affecting the borrower's credit risk, determining appropriate independent variables, so as to construct the borrower's credit risk assessment system, and carry out descriptive statistical analysis. Furthermore, two classification algorithms, decision tree and Radial-Basis Function radial basis function neural network, are used to establish the credit risk assessment model of loan borrowers in P2P network, and to carry out simulation training to evaluate the borrowers' credit effectively. Finally, through comparing the output results of the two algorithms in index selection and prediction accuracy, it is found that the four variables, interest rate, repayment period, total loan amount and the number of repayments, are the common important evaluation indexes of the two algorithms; The prediction accuracy of the decision tree model with fewer evaluation indexes is higher than that of the RBF neural network model which considers all the evaluation indexes. However, the RBF neural network model has more advantages than the RBF neural network model.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.4;F724.6

【参考文献】