论文总字数：23811字

摘 要

P2P网络借贷中借款违约可以分为借款欺诈和借款逾期，根据借款人的认证信息及债务背景等可以对借款人逾期可能以及逾期率做出预估。由于我国社会信用体系不健全、借款人主观欺诈意图感知的困难、欺诈手段不断更新和互联网的隐蔽性，目前平台在防范借款人欺诈还处于劣势一方。通过了解网贷行业发展背景，加上其在我国的成长环境的独特性，得出目前该行业的风控现状和不足，同时区分借款逾期和借款欺诈，总结出借款欺诈行为及其特点，并针对这种特点采取不同的风控措施。随着大数据技术和机器学习方法在反欺诈领域越来越受欢迎，加上借款人主观的欺诈意图必然在其某些行为中得到客观体现，因此本文对来自拍拍贷的网贷数据使用无监督学习，识别数据中的异常。通过调用python中的isolation forest模块，来对借款人在该网站上的操作记录进行学习，并得出借款人的行为属于正常类或异常类，来判断借款人是否有潜在的欺诈意图，针对存在异常行为的借款人平台可以采取相应的措施提前做好防范工作。因此，通过对借款和还款过程中借款人网站操作全过程的监控，即对贷前 贷后的双向监控，将主观欺诈意图利用客观存在反映出来，可以帮助平台主动地应对潜在的欺诈行为，做好欺诈风险防范。

关键字：网贷平台、借款欺诈、大数据

Abstract

Borrowing defaults in P2P lending can be divided into loan frauds and overdue loans. According to the borrower's identity information and debt background, the borrower's overdue and overdue rates can be estimated. Due to the unsoundness of China's social credit system, the difficulties of borrowers' subjective fraud intentions, the continuous updating of fraudulent means, and the hidden nature of the Internet, the platform is still at a disadvantage in preventing loan fraud. By understanding the background of the development of the online loan industry and the uniqueness of its growth environment in China, the current situation and deficiencies of the industry’s current wind control are deduced. At the same time, the loan overdue and loan fraud are distinguished, and the fraudulent behavior and characteristics of the loan are summarized. Different risk control measures are taken for this feature. With the increasing popularity of big data technology and machine learning methods in the field of anti-fraud, and the borrowers’ subjective fraud intentions must be objectively reflected in some of their behaviors, this paper uses data from credit loans for online loans. Unsupervised learning, identifying anomalies in the data. By calling the isolation forest module in python to learn the borrower's records of operations on the site, it is concluded that the borrower's behavior is normal or abnormal, to determine whether the borrower has potential fraud intentions. Abnormal behavior of the borrower platform can take appropriate measures to do preventive work in advance. Therefore, through the monitoring of the entire process of the borrower's website during the process of borrowing and repayment, that is, the two-way monitoring after the pre-lending and post-loan, the subjective fraud intention is reflected by the objective existence, which can help the platform to actively respond to potential frauds. , do a good job in preventing fraud risk.

Key Words: P2P lending platform, loan fraud, big data

目录

摘要 I

Abstract II

第一章 导论 4

1.1 研究背景 4

1.1.1 P2P网络借贷的概述 4

1.1.2 P2P网络借贷借款欺诈 5

1.1.3 P2P网络借贷风险控制现状 6

1.2 研究方法和思路 7

1.2.1 研究方法 7

1.2.2 研究思路与论文框架 7

1.3 本研究的现实意义 8

1.4 本研究的不足 9

第二章 文献回顾 10

2.1 P2P网络借贷的相关研究 10

2.2 贷款欺诈的相关研究 11

第三章 P2P网贷借款欺诈行为及其影响因素 13

3.1 网络借贷环境中个人借款欺诈行为 13

3.1.1 个人借款欺诈行为归纳 13

3.1.2 个人网络借款欺诈的影响因素 13

3.2 网络借贷环境中企业借款欺诈行为 15

3.2.1 企业借款欺诈行为归纳 15

3.2.2 企业网络借款欺诈的影响因素 15

第四章 实证分析 18

4.1 模型选择 18

4.2 数据来源和处理 19

4.3 结果分析 22

第五章 网贷平台欺诈风险控制策略 24

5.1 个人借款欺诈风险控制策略 24

5.2 企业借款欺诈风险控制策略 25

第六章 结论 27

致 谢 28

参考文献 29

第一章 导论

1.1 研究背景

1.1.1 P2P网络借贷的概述

从“拍拍贷”在2007年创立至今，P2P网贷在我国从最初的缓慢发展慢慢演变到爆炸增长，后来随着监管框架的建立，目前的P2P网贷正在监管下成长。在此期间出现了多种商业运营模式，大致可归纳为有担保有抵押模式、有担保无抵押模式和无担保无抵押模式三类^[1]。通过查阅网贷之家从2014年至2017年的行业年报（表1-1），截止2017年年底，我国正常运营的网贷平台数量有1931家，而截止2018年4月历史累计停业及问题平台数量为4198家。

表1-1 网贷之家2014-2017中国网贷行业年报

剩余内容已隐藏，请支付后下载全文，论文总字数：23811字