基于LightGBM算法的P2P项目信用评级模型的设计及应用
|
摘要
研究目标:在大数据和互联网金融发展的背景下,依据个人信用,有效控制P2P项目的违约率以保证相关金融项目或平台的良好运营。研究方法:本文基于美国P2P平台Lending Club的海量真实交易数据,采用"多观测"与"多维度"两种数据清洗方式,运用2016年微软亚洲研究院提出的机器学习算法LightGBM,兼顾权威性和创新性地对平台内贷款项目的违约风险进行预测,并对不同数据清洗方法的结果进行比较。研究发现:基于多观测的LightGBM算法的预测结果最好,比Lending Club平台历史交易数据算的平均履约率提升了1.28个百分点,可减少约1.17亿美元的违约借款。研究创新:运用不同的数据清洗方式和较为前沿的机器学习算法(LightGBM)预测违约率。研究价值:在LightGBM算法得出违约率影响因素的结果基础上,可以明确Lending Club及广大P2P平台的改进内容和各国在该领域内发展完善的方向。
Research Objectives:In the context of big data and Internet finance development,it is effective to control the default rate of P2 Pprojects to ensure good operation of relevant financial projects or platforms according to personal credit.Research Methods:In this paper,based on the P2 Pplatform Lending Club massive real transaction data,we use the‘multi-observation'and ‘multi-dimensional'two kinds of data cleaning method,and predict the risk of default by the 2016 Asian Microsoft LightGBM machine learning algorithms of authority and innovative.Then compare the results of different data cleaning method.Research Findings:LightGBM algorithm based on multi-observations of predicted results is the best.Lending Club platform historical trading data to calculate the average execution rate of 1.28%,can reduce about MYM117 million in loan default.Research Innovations:Using different data cleaning methods and more advanced machine learning algorithm(LightGBM)to predict default rate.Research Value:Based on the result of the influencing factors of default rate,not only the suggestions on development of Lending club and P2 Pplatforms are pointed out,but also the direction of national development in this field.
Research Objectives:In the context of big data and Internet finance development,it is effective to control the default rate of P2 Pprojects to ensure good operation of relevant financial projects or platforms according to personal credit.Research Methods:In this paper,based on the P2 Pplatform Lending Club massive real transaction data,we use the‘multi-observation'and ‘multi-dimensional'two kinds of data cleaning method,and predict the risk of default by the 2016 Asian Microsoft LightGBM machine learning algorithms of authority and innovative.Then compare the results of different data cleaning method.Research Findings:LightGBM algorithm based on multi-observations of predicted results is the best.Lending Club platform historical trading data to calculate the average execution rate of 1.28%,can reduce about MYM117 million in loan default.Research Innovations:Using different data cleaning methods and more advanced machine learning algorithm(LightGBM)to predict default rate.Research Value:Based on the result of the influencing factors of default rate,not only the suggestions on development of Lending club and P2 Pplatforms are pointed out,but also the direction of national development in this field.
引文
[1]Arya S.,Eckel C.C.,Wichman C.,2013,Anatomy of the Credit Score[J],Journal of Economic Behavior and Organization,95,175~185.
[2]Berger S.C.,Gleisner F.,2009,Emergence of Financial Intermediaries in Electronic Markets:The Case of Online P2P Lending[J],Business Research,2(1),39~65.
[3]Everett C.R.,2015,Group Membership,Relationship Banking and Loan Default Risk:The Case of Online Social Lending[J],Banking and Finance Review,7(2),15~54.
[4]Freedman S.,Jin G.Z.,2008,Do Social Networks Solve Information Problems for Peer-to-Peer Lending?Evidence from Prosper.Com[R],NET Institute Working Paper.
[5]Freedman S.,Jin G.Z.,2011,Learning by Doing with Asymmetric Information:Evidence from Prosper.Com[R],NBER Working Paper No.16855.
[6]Li H.,Zhang Y.,Zhang N.,Jia H.,2016,Detecting the Abnormal Lenders from P2P Lending Data[J],Procedia Computer Science,91,357~361.
[7]Herzenstein M.,Andrews R.L.,Dholakia U.,Lyandres E.,2008,The Democratization of Personal Consumer Loans?Determinants of Success in Online Peer-to-Peer Lending Communities[R],Boston University School of Management Research Paper.
[8]Herzenstein M.,Sonenshein S.,Dholakia U.M.,2011,Tell Me a Good Story and I May Lend You Money:The Role of Narratives in Peer-to-Peer Lending Decisions[J],Journal of Marketing Research,48,138~149.
[9]Iyer R.,Khwaja A.I.,Luttmer E.F.,Shue K.,2016,Screening Peers Softly:Inferring the Quality of Small Borrowers[J],Management Science,62(6),1554~1577.
[10]Kupp M.,Anderson J.,2007,ZOPA:Web 2.0 Meets Retail Banking[J],Business Strategy Review,18(3),11~17.
[11]Klafft M.,2008,Peer to Peer Lending:Auctioning Microcredits over the Internet[A].in Agarwal A.,Khurana R.(eds),Proceedings of the International Conference on Information Systems[C],Technology and Management,IMT,Dubai.
[12]Lin M.,2009,Peer-to-Peer Lending:An Empirical Study[C],AMCIS 2009Donsortium.
[13]Lin M.,Prabhala N,R,,Viswanathan S.,2011,Judging Borrowers by the Company They Keep:Social Networks and Adverse Selection in Online Peer-to-peer Lending[C],Western Finance Association 2009Annual Meeting Paper.
[14]Malekipirbazari M.,Aksakalli V.,2015,Risk Assessment in Social Lending via Random Forests[J],Expert Systems With Applications,42(10),4621~4631.
[15]Meng Q.,Ke G.,Wang T.,Chen W.,Ye Q.,Ma Z.,Liu T.,2016,A Communication-Efficient Parallel Algorithm for Decision Tree[C],30th Conference on Neural Information Processing Systems(NIPS 2016).
[16]Ravina E.,2007,Beauty,Personal Characteristics and Trust in Credit Markets[C],American Law&Economics Association Annual Meetings.
[17]Renton P.,2012,The Lending Club Story[M],Renton Media LLC.
[18]艾金娣:《P2P网络借贷平台风险防范》[J],《中国金融》2012年第14期。
[19]巴曙松、侯畅、唐时达:《大数据风控的现状、问题及优化路径》[J],《金融理论与实践》2016年第2期。
[20]李焰、高弋君、李珍妮、才子豪、王冰婷、杨宇轩:《借款人描述性信息对投资人决策的影响---基于P2P网络借贷平台的分析》[J],《经济研究》2014年第S1期。
[21]李悦雷、郭阳、张维:《中国P2P小额贷款市场借贷成功率影响因素分析》[J],《金融研究》2013年第7期。
[22]廖理、张伟强:《P2P网络借贷实证研究:一个文献综述》[J],《清华大学学报(哲学社会科学版)》2017年第2期。
[23]廖理、李梦然、王正位:《聪明的投资者:非完全市场化利率与风险识别---来自P2P网络借贷的证据》[J],《经济研究》2014年第7期。
[24]廖理、吉霖、张伟强:《借贷市场能准确识别学历的价值吗?---来自P2P平台的经验证据》[J],《金融研究》2015年第3期。
[25]卢馨、李慧敏:《P2P网络借贷的运行模式与风险管控》[J],《改革》2015年第2期。
[26]马若微、唐春阳:《考虑误判损失的Logistic违约预测模型构建》[J],《系统工程理论与实践》2007年第8期。
[27]王会娟、廖理:《中国P2P网络借贷平台信用认证机制研究---来自“人人贷”的经验证据》[J],《中国工业经济》2014年第4期。
[28]微软亚洲研究院:《开源|LightGBM:三天内收获GitHub 1000+星》[OL/EB],http://www.sohu.com/a/123480446_133098.
[29]张昊、纪宏超、张红宇:《XGBoost算法在电子商务商品推荐中的应用》[J],《物联网技术》2017年第2期。
[2]Berger S.C.,Gleisner F.,2009,Emergence of Financial Intermediaries in Electronic Markets:The Case of Online P2P Lending[J],Business Research,2(1),39~65.
[3]Everett C.R.,2015,Group Membership,Relationship Banking and Loan Default Risk:The Case of Online Social Lending[J],Banking and Finance Review,7(2),15~54.
[4]Freedman S.,Jin G.Z.,2008,Do Social Networks Solve Information Problems for Peer-to-Peer Lending?Evidence from Prosper.Com[R],NET Institute Working Paper.
[5]Freedman S.,Jin G.Z.,2011,Learning by Doing with Asymmetric Information:Evidence from Prosper.Com[R],NBER Working Paper No.16855.
[6]Li H.,Zhang Y.,Zhang N.,Jia H.,2016,Detecting the Abnormal Lenders from P2P Lending Data[J],Procedia Computer Science,91,357~361.
[7]Herzenstein M.,Andrews R.L.,Dholakia U.,Lyandres E.,2008,The Democratization of Personal Consumer Loans?Determinants of Success in Online Peer-to-Peer Lending Communities[R],Boston University School of Management Research Paper.
[8]Herzenstein M.,Sonenshein S.,Dholakia U.M.,2011,Tell Me a Good Story and I May Lend You Money:The Role of Narratives in Peer-to-Peer Lending Decisions[J],Journal of Marketing Research,48,138~149.
[9]Iyer R.,Khwaja A.I.,Luttmer E.F.,Shue K.,2016,Screening Peers Softly:Inferring the Quality of Small Borrowers[J],Management Science,62(6),1554~1577.
[10]Kupp M.,Anderson J.,2007,ZOPA:Web 2.0 Meets Retail Banking[J],Business Strategy Review,18(3),11~17.
[11]Klafft M.,2008,Peer to Peer Lending:Auctioning Microcredits over the Internet[A].in Agarwal A.,Khurana R.(eds),Proceedings of the International Conference on Information Systems[C],Technology and Management,IMT,Dubai.
[12]Lin M.,2009,Peer-to-Peer Lending:An Empirical Study[C],AMCIS 2009Donsortium.
[13]Lin M.,Prabhala N,R,,Viswanathan S.,2011,Judging Borrowers by the Company They Keep:Social Networks and Adverse Selection in Online Peer-to-peer Lending[C],Western Finance Association 2009Annual Meeting Paper.
[14]Malekipirbazari M.,Aksakalli V.,2015,Risk Assessment in Social Lending via Random Forests[J],Expert Systems With Applications,42(10),4621~4631.
[15]Meng Q.,Ke G.,Wang T.,Chen W.,Ye Q.,Ma Z.,Liu T.,2016,A Communication-Efficient Parallel Algorithm for Decision Tree[C],30th Conference on Neural Information Processing Systems(NIPS 2016).
[16]Ravina E.,2007,Beauty,Personal Characteristics and Trust in Credit Markets[C],American Law&Economics Association Annual Meetings.
[17]Renton P.,2012,The Lending Club Story[M],Renton Media LLC.
[18]艾金娣:《P2P网络借贷平台风险防范》[J],《中国金融》2012年第14期。
[19]巴曙松、侯畅、唐时达:《大数据风控的现状、问题及优化路径》[J],《金融理论与实践》2016年第2期。
[20]李焰、高弋君、李珍妮、才子豪、王冰婷、杨宇轩:《借款人描述性信息对投资人决策的影响---基于P2P网络借贷平台的分析》[J],《经济研究》2014年第S1期。
[21]李悦雷、郭阳、张维:《中国P2P小额贷款市场借贷成功率影响因素分析》[J],《金融研究》2013年第7期。
[22]廖理、张伟强:《P2P网络借贷实证研究:一个文献综述》[J],《清华大学学报(哲学社会科学版)》2017年第2期。
[23]廖理、李梦然、王正位:《聪明的投资者:非完全市场化利率与风险识别---来自P2P网络借贷的证据》[J],《经济研究》2014年第7期。
[24]廖理、吉霖、张伟强:《借贷市场能准确识别学历的价值吗?---来自P2P平台的经验证据》[J],《金融研究》2015年第3期。
[25]卢馨、李慧敏:《P2P网络借贷的运行模式与风险管控》[J],《改革》2015年第2期。
[26]马若微、唐春阳:《考虑误判损失的Logistic违约预测模型构建》[J],《系统工程理论与实践》2007年第8期。
[27]王会娟、廖理:《中国P2P网络借贷平台信用认证机制研究---来自“人人贷”的经验证据》[J],《中国工业经济》2014年第4期。
[28]微软亚洲研究院:《开源|LightGBM:三天内收获GitHub 1000+星》[OL/EB],http://www.sohu.com/a/123480446_133098.
[29]张昊、纪宏超、张红宇:《XGBoost算法在电子商务商品推荐中的应用》[J],《物联网技术》2017年第2期。