基于客户细分的大学生网贷项目信用风险的识别与度量

发布时间：2018-01-11 15:35

本文关键词：基于客户细分的大学生网贷项目信用风险的识别与度量　出处：《上海师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：P2P小额借贷模式近几年来发展迅速,截至2016年11月底,我国正常运营的P2P网贷平台数量高达2534家,行业竞争愈演愈烈,行业市场也开始出现细分。在众多的细分市场中,大学生P2P网贷发展迅猛,独树一帜。大学生相对于其他网贷借款人,有着身份易认证、违约成本高、父母为其潜在担保等优势,信用风险相对较小。但是,由于网贷简单便捷导致冲动消费、大学生没有固定的经济来源、征信意识不强等原因,大学生网贷平台上仍存在大量的逾期违约现象。本文主要是策划与实施一个大学生网贷项目信用风险识别与度量方案。本方案实施与验证均是基于“速溶360”平台项目信息。本方案实施主要包含4个步骤:首先,基于客户细分理论,分别从新老客户、借款者学历两个方面对大学生网贷项目进行分类,利用两步聚类法根据借款者学历将人群分为了3类:高学历人群、普通人群以及低学历人群;其次,针对每个类群项目,利用特征选择方法,进行降维处理,提高了后续方法的准确性与稳定性;然后,利用XGBoost算法,对每个类群的网贷项目信用风险进行识别,结果显示,该算法在信用风险识别上的运用效果很好,准确率很高;最后,利用XGBoost模型结果与评分卡模型对网贷项目的信用风险进行度量并检验结果的有效性。本方案的实施效果很好,符合预期,是一个较好的信用风险识别方案。此外,本方案的设计与实施,不仅为网贷项目信用风险识别提供了一种新的思路,也验证了 XGBoost模型在信用风险识别问题上的有效性。
[Abstract]:Peer-to-peer microfinance model has developed rapidly in recent years. By the end of November 2016, the number of P2P network lending platforms in China is as high as 2534, and the competition in the industry is becoming more and more intense. Industry market also began to subdivide. In many segments of the market, P2P network loans for college students developed rapidly, unique. Compared with other network loan borrowers, college students have easy identity authentication, high cost of breach of contract. Parents for its potential guarantee and other advantages, credit risk is relatively small. However, because of the simple and convenient Internet loans led to impulse consumption, college students do not have a fixed source of financial resources, credit awareness is not strong and other reasons. There is still a large number of overdue breach of contract on the university student network loan platform. This paper mainly plans and implements a scheme to identify and measure the credit risk of the university student network loan project. The implementation and verification of this scheme are all based on " Instant 360 "platform project information. The implementation of this scheme mainly includes four steps: first. Based on the theory of customer segmentation, this paper classifies college students' online loan projects from two aspects: new and old customers, borrowers' qualifications, and classifies the population into three categories according to the borrower's degree by using two-step clustering method: the high-educated group. The general population and the low-educated population; Secondly, the method of feature selection is used to reduce the dimension of each group item, which improves the accuracy and stability of the follow-up method. Then, the XGBoost algorithm is used to identify the credit risk of each network loan project. The results show that the algorithm is effective and accurate in the identification of credit risk. Finally, using the XGBoost model results and scoring card model to measure the credit risk of network loan projects and test the effectiveness of the results. The implementation of this scheme is very good, in line with expectations. In addition, the design and implementation of this scheme not only provides a new way to identify the credit risk of the network loan project. The validity of XGBoost model in credit risk identification is also verified.
【学位授予单位】：上海师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：G645.5;F724.6;F832.4

【参考文献】