基于机器学习的P2P网络借贷平台风险监测与预警  

Early Risk Identification and Warning of P2P Online Lending Platform

作　　者：李爽 刘娅娅 李晓彤 LI Shuang;LIU Ya-ya;LI Xiao-tong(School of Statistics,Xi’an University of Finance and Economics,Xi’an 710100,China)

机构地区：[1]西安财经大学统计学院,陕西西安710100

出　　处：《西安财经大学学报》2022年第2期62-73,共12页Journal of Xi’an University of Finance and Economics

基　　金：国家社会科学基金项目“基于电商平台数据的网购人群消费特征统计分析研究”(17BTJ022)。

摘　　要：自2015年底“e租宝”事件发生以来,网络借贷平台问题层出不穷。P2P网络借贷市场已经成为互联网金融风险的重要聚集地,不仅扰乱了我国金融市场的秩序,而且造成了网络借贷平台从业者失业、投资者利益受损。对P2P网络借贷平台存在的风险进行识别是避免问题平台出现的关键环节。文章借鉴我国商业银行对风险评价的“三性”原则和LAPP法,构建P2P网络借贷平台的风险评价指标体系;基于网贷之家2019年10月144家正常运营平台的数据,利用高斯混合模型识别平台风险等级,将平台风险类型划分为重警、中警、轻警、微警、无警等五个风险等级;运用四种机器学习方法研究了P2P网络借贷平台风险。结果表明:朴素贝叶斯模型对风险的识别能力最强,流动性风险和信用风险是两个重要的风险维度。本研究对网贷平台风险的识别与预警,有助于投资者理性投资、政府高效监管以及平台管理者提前了解自身运营状况。Since the“E-zubao”incident at the end of 2015,the problems of online lending platforms have emerged in an endless stream.The P2P online lending market has become the main place of internet financial risks,which not only disrupts the order of China’s financial market,but also causes unemployment and interests damage.The key to avoiding the problem platform is to identify the risks of the P2P network lending platform.This article draws on the“three natures”principle of risk assessment of Chinese commercial banks to build a risk assessment system for P2P online lending platforms;based on the data of 144 normal operating platforms of home of online loan in October 2019.The Gaussian mixture model is used to identify the risk levels of the platform,and the risk types are divided into five risk levels:heavy alarm,medium alarm,light alarm,micro alarm,and no alarm.Four machine learning methods are used to study the risks of P2P online lending platform.The results show that the Naive Bayes model has the strongest ability to identify risks.Liquidity risk and credit risk are two important risk dimensions.The identification and warning of the risks of the online lending platform will help investors to make rational investments,government efficient supervision,and platform managers to understand their own operations in advance.

关 键 词：P2P网络借贷平台 风险等级 风险预警 高斯混合模型 机器学习 

分 类 号：F062.2[经济管理—政治经济学]