P2P网络借贷满标概率预测研究

发布时间：2018-07-05 20:32

本文选题：P2P网络借贷 + 借款满标　；参考：《河南大学》2017年硕士论文

【摘要】：随着互联网安全、第三方支付技术的日渐成熟、人们日常生活中对网上交易信任的增加,互联网金融行业迅速发展。P2P网络借贷作为典型的互联网金融模式,为急需资金的个人和中小企业提供了融资渠道,其依靠第三方网络平台,为个人与个人之间提供了公开透明的小额信用交易的可能,也为实现金融脱媒和践行普惠金融做出了有益贡献。网络借贷自2005年兴起以来,在全球范围内发展迅速,但也面临着借贷成功率较低、资金成本过高等问题。借款满标是指借款人在网贷平台规定的期限内其列表可得到足够的出借人关注和投标,进而足额筹集到所需资金的状态。本文通过对国内外文献的梳理和分析,试图从借款人信息出发,探讨投资人通过第三方网贷平台进行投资决策的过程,为借款人提高借款满标概率及投资者提高投资精度提供有益建议。借贷市场中的信息传播方式及公开程度对金融机构的发展至关重要,在P2P网贷市场中交易双方互不相识仅通过网络产生联系,因此借款人在关键信息方面通常更有优势,大量潜在真实信息常常由借款人掌握,投资人则处于信息劣势,借贷双方存在着明显的信息不对称。本文将信息不对称理论应用到网贷行业,从网络贷款平台运营现状入手,审视网贷经营模式,重现P2P网贷交易流程,以阐明借款人信息与借款满标之间的关系,并构建借款满标概率预测模型。文中对国内借贷网站——拍拍贷的借贷机制进行了着重介绍,作为首家纯信用无担保网贷平台,拍拍贷从性质上讲最接近P2P网络借贷的本质,因此选择该平台作为研究主体,同时也从该平台运用网络爬虫软件选取实证分析的数据来源。文章首先对选取的数据进行聚类分析,为了解不同特征的借款人使用平台的效率,文章选取5个维度的指标,对全部样本数据进行聚类分析,分析结果将8个信用等级的借款人分为四类。从而得出评估授信状况相似的借款人行为模式是有差异的,而信用等级不同的借款人之间也会存在共性,也即信用等级部分反映了借款人的违约风险进而影响其借贷成功可能性。因此对于借款人的动态行为规律,只以信用等级来界定是有失偏颇的。因此挖掘出信用等级这一信号所不能显示的信息,识别用户在平台的表现,可以找出借款人中出现的不同的行为模式,同时为借款满标概率的研究提供了契机。希望经过借款满标概率预测模型的构建,可以借助借款人在网贷平台上披露的信息预测其借款满标可能性。接着,文章选取指标以构建借款满标概率预测模型,经过对样本数据进行预处理及多重共线性诊断,得出自变量之间存在多重共线性。本文为解决解释变量间存在的多重共线性问题,以使最终构建模型更加精确,在阅读相关文献后选择使用主成分改进的逻辑回归方法。这种方法的核心在于经过变换可将原本具有相关性的解释变量综合为少数几个综合指标,提取出的少数综合指标能反映原来多个变量的信息。最终以借款人信息为自变量,以借贷满标与否为因变量,依据抓取的数据,经过数据处理,用提取出的主成分代替原有全部解释变量进行逻辑回归,用二元逻辑回归方法构建借款满标概率预测模型,之后再对各提取出的主成分进行还原,来研究其他若干变量对满标概率的作用方向和影响程度。为借款人提高借款满标概率、投资人优化投资决策提供参考。最后针对当前借款成功率较低,信用机制不健全等现状,提出相应优化对策及对未来的展望,并主要从完善社会征信体系、加强互联网金融领域立法、健全网贷行业风险应对机制三个方面给出建议。
[Abstract]:With the security of the Internet and the growing maturity of the third party payment technology and the increasing trust in online transactions in people's daily life, the Internet finance industry has rapidly developed.P2P network lending as a typical internet financial model, providing financing channels for individuals and small and medium enterprises, which are in urgent need of funds, and rely on the third party network platform as individuals. The possibility of open and transparent small credit transactions with individuals has also contributed to the realization of financial disintermediation and the implementation of Inclusive Finance. Since the rise of 2005, Internet lending has developed rapidly around the world, but it also faces the low success rate of borrowing and the higher funding. The loan full scale refers to the borrower in the net. In the time limit set by the loan platform, the list of the borrowers can get enough attention and bid, and then raise the state of the required funds in full. By combing and analyzing the literature at home and abroad, this paper tries to discuss the process of investment decision by the investor through the third party net loan platform to improve the borrower's full loan. The way and openness of information dissemination in the loan market is very important to the development of financial institutions. In the P2P net loan market, the non acquaintances of the two parties are only connected through the network, so the borrower is usually more advantageous in the key information and a large number of potential real information. In this paper, the information asymmetry theory is applied to the net loan industry. This paper applies the information asymmetry theory to the net loan industry, starts with the operation status of the network loan platform, examines the operation mode of the net loan, reproduces the P2P net loan transaction flow, in order to clarify the relationship between the borrower's information and the full standard of the loan. In this paper, the loan full standard prediction model is built and the loan mechanism of the domestic loan website - pat loan is introduced in this paper. As the first platform of the first pure credit unsecured net loan, the patting loan is most close to the nature of the P2P network loan from the nature, so the platform is chosen as the research subject and the network is also used on the platform. In order to understand the efficiency of the borrowers with different features, the article selects 5 dimensions to analyze the efficiency of the different features of the borrowers, and analyzes the data of all the samples and divides the 8 credit rating borrowers into four categories. The behavior patterns of borrowers with similar status are different, and the borrowers with different credit grades also have commonality, that is, the credit grade reflects the borrower's default risk and then affects the possibility of borrowing success. Therefore, the dynamic behavior law of the borrower is biased by the credit rating. It can identify the information that the credit rating can not show, identify the user's performance on the platform, find out the different behavior patterns in the borrower, and provide an opportunity for the study of the probability of the loan full scale. The information forecast its loan full scale possibility. Then, the article selects the index to construct the probability prediction model of the loan full scale, through the preprocessing of the sample data and the multiple collinear diagnosis, the multiple collinearity exists between the independent variables. This paper is to solve the multiply collinearity problem between the explanatory variables, so as to make the final construction model more The core of this method is to integrate the original explanatory variables with the original correlation into a few comprehensive indexes, and the extracted minority index can reflect the original variable quantity information. Finally, the borrower information is the independent variable, Taking the full standard of borrowing and lending as the dependent variable, according to the captured data, after data processing, the extracted principal component is used instead of all the original explanatory variables to carry out logic regression. The two element logic regression method is used to construct the loan full scale probability prediction model, then the main components of each extraction are reduced to study the full scale of the other variables. The direction and extent of the effect of the probability. It provides a reference for the borrower to improve the loan full standard probability and the investor to optimize the investment decision. Finally, it puts forward the corresponding optimization countermeasures and prospects for the future, aiming at the low success rate of the current borrowing and the imperfect credit mechanism and so on, and mainly from the good social credit system to strengthen the Internet financial field. Three suggestions are given to improve the risk response mechanism of the net loan industry.
【学位授予单位】：河南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.6;F832.4

【参考文献】