P2P随着互联网金融的发展迅速兴起，但极高的违约率让投资者付出了巨额代价。已有的风险控制策略存在信用评估和违约鉴别不准确，投资决策盲目等缺点。本文从投资者的视角出发，针对如何选取合适的借贷项目以及如何合理分配投资资金的问题展开研究。以Lending Club为实验对象，分别从二分类和多分类两个角度进行网贷风险控制。
第一，构建了一个信息非冗余且具有显著违约鉴别能力的信用评价指标体系，为网络借贷违约鉴别做准备。根据信用5C分析法确立了信用评价指标体系的一级指标层，将IV-WOE框架与Spearman-Boruta算法相结合，进行了两阶段指标筛选。选出36个指标确立了最终信用评价指标体系，并给出了其与信用5C标准的对应关系。这不仅避免了人为主观误删的问题，也保证了选中指标具有较强的违约鉴别能力。
第二，从二分类角度进行网络借贷投资风险控制，提出了基于Stacking集成学习的网络借贷违约鉴别模型。按照违约样本比划分训练集和测试集，训练了9种单一分类器，选取4种效果较优的分类器作为基分类器。并以ROC-AUC为性能评价标准，采用5则交叉验证训练300次进行随机搜索调参。最终构建了以ANN、RF、AdaBoost和XGBoost为基分类器，LR为元分类器的Stacking集成分类模型，实现了网络借贷违约鉴别。对比结果表明：Stacking集成分类器融合了各单一分类器的优势，具有更强的违约识别能力，可辅助投资者选取合适的借贷项目，是更能满足网络借贷业务需求的违约鉴别模型。
第三，从多分类角度进行网络借贷投资风险控制，提出了基于多分类期望收益率矩阵的投资组合改进模型。将网贷中的信用误分损失考虑进来，构建了多分类的期望收益率矩阵。使用违约概率差值衡量借贷的接近程度，根据高斯核函数确定权重，对相似历史借贷进行加权平均，量化了新借贷的预期收益率和风险率。最终将网络借贷投资转化为一个二次规划问题，提出了投资组合改进模型。并基于Lending Club设计了六组实验，验证了投资组合改进模型在网贷中的风控效果。结果表明：针对不同的参数设置，投资组合改进模型优化后的风险率始终小于原本应承担的风险率。该模型在保证最小收益率的同时，可辅助投资者合理分配投资资金，在网络借贷投资风险控制领域具有实用价值。

P2P is rising rapidly with the development of Internet finance, but the extremely high default rate has made investors pay a huge price. The existing risk control strategies have shortcomings such as inaccurate credit assessment and default identification, and blind investment decisions. From the perspective of investors, this thesis studies how to select appropriate loan projects and how to allocate investment funds rationally. Using the data of Lending Club as the experimental object, risk control is carried out from the perspectives of binary classifications and multiple classifications respectively. 

Firstly, A credit evaluation index system with non redundant information and significant default identification ability is constructed, which is to prepare for the default identification of P2P lending. According to the credit 5C analysis method, the first level index layer of the credit evaluation index system is established, and the two-stage index screening is carried out by combining the IV-WOE framework with Spearman-Boruta algorithm. 36 indexes are selected, the final credit evaluation index system is established, and its corresponding relationship with the credit 5C standard is given. This not only avoids the problem of artificial subjective deletion, but also ensures that the selected indicators have strong ability of default identification.

Secondly, from the perspective of binary classification, the risk control of online lending investment is carried out, the default identification model of online lending based on Stacking ensemble learning is proposed. According to the default sample ratio, the training set and testing set are divided, nine single classifiers are trained, and four better classifiers are selected as the base classifier. Taking ROC-AUC as the performance evaluation standard, five cross validation training were used for 300 times, and random search was carried out to adjust the parameters. Finally, with ANN, RF, AdaBoost and XGBoost as the base classifier, LR as the meta-classifier, the Stacking ensemble classification model is constructed to realize the default identification of online lending. The comparison results show that the Stacking ensemble 。classifier combines the advantages of each single classifier, has a stronger default identification ability, can assist investors in selecting suitable lending projects, so it is a default identification model that can better meet the needs of online lending business.

Finally, from the perspective of multiple classification, the risk control of online lending investment is carried out, an improved portfolio model based on multi classification expected rate of return matrix is proposed. Considering the credit misclassification loss in lending, a multi classification expected rate of return matrix is constructed. The default probability difference is used to measure the proximity of loans, the weight is determined according to the Gaussian kernel function, the weighted average of similar historical loans is carried out, and the expected return and risk rate of new loans are quantified. Finally, the online loan investment is transformed into a quadratic programming problem, and an improved portfolio model is proposed. Six groups of experiments are designed based on Lending Club to verify the risk control effect of the improved portfolio model in online lending. The results show that for different parameter settings, the optimized risk rate of the improved portfolio model is always less than the original risk rate. The model can not only ensure the minimum rate of return, but also assist investors to allocate investment funds reasonably, which has practical value in the field of risk control of online lending investment.

[1]Su C W, Qin M, Tao R, Umar M. Financial Implications of Fourth Industrial Revolution: Can Bitcoin Improve Prospects of Energy Investment?[J]. Technological Forecasting and Social Change, 2020, 158: 120-178.

[2]Xu M, David J M, Kim S H. The Fourth Industrial Revolution: Opportunities and Challenges[J]. Internationa Journal of Financial Research, 2018, 9: 90-95.

[3]Loubere N. China’s Internet Finance Boom and Tyrannies of Inclusion. China Perspect[J]. China Perspectives, 2017, 4: 9-18.

[4]廖先玲，蔡成铭. 我国互联网金融风险研究的网络结构和主题聚类——基于CiteSpace的知识图谱分析[J]. 金融理论与实践，2018，(8): 71-76.

[5]Lin M, Prabhala N R, Viswanathan S. Judging Borrowers by the Company They Keep: Friendship Networks and Information Asymmetry in Online Peer-to-Peer Lending[J]. Management Science, 2013, 59(1): 17-35.

[6]Chen D, Lai F, Lin Z. A Trust Model for Online Peer-to-Peer Lending: a Lender’s Perspective[J]. Information Technology and Management, 2014, 15: 239-254.

[7]王建刚. 校园网络借贷的风险与规制[J]. 中国商论，2017，(3): 34-37.

[8]曹一梅. 我国P2P网络借贷平台竞争力研究[D]. 西安：西安电子科技大学，2017.

[9]庄雅玲. 个人特质与P2P平台融资行为的关系研究[D]. 哈尔滨：哈尔滨工业大学，2017.

[10]庄越融. 基于演化博弈的P2P网络借贷平台激励监管研究[D]. 杭州：浙江财经大学，2018.

[11]高郡. 中美P2P网络借贷违约率影响因素研究 ——以人人贷和Lending Club为例[D]. 武汉：武汉科技大学，2018.

[12]隋昕. 基于Logit模型的P2P网络借贷平台借款人信用风险影响因素研究[D]. 哈尔滨：哈尔滨商业大学，2017.

[13]魏明侠，翟梦凡. 网贷市场出借意愿:基于拍拍贷用户数据的实证研究[J]. 经济与管理评论，2017, 33(4): 87-93.

[14]蓝波. P2P网络借贷平台的运行机理及其匹配效率的实证研究[D]. 南京：东南大学，2016.

[15]李伟. 数据驱动的消费金融违约风险预测方法研究[D]. 合肥：合肥工业大学，2019.

[16]杨献. 网络借贷平台的信息披露及协同治理研究[D]. 南京：南京邮电大学，2019.

[17]Herzenstein M, Dholakia U M, Andrews R L. Strategic Herding Behavior in Peer-to-Peer Loan Auctions[J]. Journal of Interactive Marketing, 2011, 25(1):27-36.

[18]万洁. 商业银行发展互联网金融业务的战略研究——以中国银行为例[D]. 武汉：南昌大学，2018.

[19]Iyer R., Khwaja A I, Luttmer E F P, Shue K. Screening Peers Softly: Inferring the Quality of Small Borrowers[J]. Management Science, 2016, 62(6): 1533-1841.

[20]Pope D G, Sydnor J R. Geographic Variation in the Gender Differences in Test Scores[J]. Journal of Economic Perspectives, 2010, 24(2): 95-108.

[21]Lin M., Prabhala N R, Viswanathan S. Judging Borrowers by the Company They Keep: Friendship Networks and Information Asymmetry in Online Peer-to-Peer Lending[J]. Management Science, 2013, 59(1): 17-35.

[22]González L, Loureiro Y K. When Can a Photo Increase Credit? The Impact of Lender and Borrower Profiles on Online P2P Loans[J]. Journal of Behavioral and Experimental Finance, 2014, 2: 44-58.

[23]Lin X, Li X, Zheng Z. Evaluating Borrower’s Default Risk in Peer-to-Peer Lending:Evidence from a Lending Platform in China. Applied Economics[J]. Applied Economics, 2017, 49(35): 3538-3545.

[24]冯素玲，赵家玲，赵书. 女性借款人对降低网贷市场违约风险有积极效应吗?——来自“拍拍贷”的实证研究[J]. 济南大学学报（社会科学版），2020, 30(2): 93-101.

[25]Jagric V, Kracun D, Jagric T. Does Non-Linearity Matter in Retail Credit Risk Modeling?[J]. Finance a Uver: Czech Journal of Economics & Finance, 2011, 61(4): 384-402.

[26]张洪祥，毛志忠．基于多维时间序列的灰色模糊信用评价研究[J]． 管理科学学报，2011，14(1): 29-37.

[27]张发明． 一种融合SOM与K-means算法的动态信用评价方法及应用[J]. 运筹与管理，2014，23(6): 187-192.

[28]Yuan Y, Yang Y, Wang H Q. Research on Credit Risk Assessment of P2P Online Lending Platform[J]. Management Science and Engineering, 2020, 9(3): 157-164.

[29]Malekipirbazari M, Aksakalli V, Risk Assessment in Social Lending via Random Forests[J]. Expert Systems with Applications, 2015, 42(10): 4621-4631.

[30]Wang G, Hao J, Ma J, Jiang H. A Comparative Assessment of Ensemble Learning for Credit Scoring[J]. Expert Systems with Applications, 2011, 38(1): 223-230.

[31]Crook J N, Edelman D B, Thomas L C. Recent Developments in Consumer Credit Risk Assessment[J]. European Journal of Operational Research, 2007, 183(3): 1447-1465.

[32]Ngai E W T, Hu Y, Wong Y H, Chen Y, Sun X. The Application of Data Mining Techniques in Financial Fraud Detection: A Classification Framework and an Academic Review of Literature[J]. Decision Support Systems, 2011, 50(3): 559-569.

[33]Emekter R, Tu Y, Jirasakuldech B, Lu M. Evaluating Credit Risk and Loan Performance in Online Peer-to-Peer (P2P) Lending[J]. Applied Economics, 2015, 47(1): 54-70.

[34]Kruppa J, Ziegler A, König I R. Risk Estimation and Risk Prediction Using Machine-Learning Methods[J]. Human Genetics, 2012, 131: 1639-1654.

[35]Baesens B, Van Gestel T, Viaene S, Stepanova M, Suykens J, Vanthienen J. Benchmarking State-of-the-Art Classification Algorithms for Credit Scoring[J]. Journal of the Operational Research Society, 2003, 54: 627-635.

[36]Lessmann S, Baesens B, Seow H V, Thomas L C. Benchmarking State-Ofthe-Art Classification Algorithms for Credit Scoring: An Update of Research[J]. European Journal of Operational Research, 2015, 247(1): 124-136.

[37]Bolton R J, Hand D J. Statistical Fraud Detection: A Review[J], Statistical Science, 2002, 17(3): 235-255.

[38]Bhattacharyya S, Jha S, Tharakunnel K, Westland J C. Data Mining for Credit Card Fraud: A Comparative Study[J]. Decision Support Systems, 2011, 50 (3), 602-613.

[39]Harris T, Credit Scoring Using the Clustered Support Vector Machine[J]. Expert Systems with Applications, 2015, 42(2): 741-750.

[40]Berger A N, Frame W S, Ioannidou V. Reexamining the Empirical Relation Between Loan Risk and Collateral: the Roles of Collateral Liquidity and Types[J]. Journal of Financial Intermediation, 2016, 26, 28-46.

[41]Zhu Y, Xie C, Sun B, Wang G J, Yan X G. Predicting China’s SME Credit Risk in Supply Chain Financing by Logistic Regression, Artificial Neural Network and Hybrid Models[J]. Sustainability, 2016, 8(5): 433.

[42]Zhang Y L, Zhou J, Zheng W, Feng J, Li L, Liu Z, Li M, Zhang Z, Chen C, Li X, Qi Y, Zhou Z H. Distributed Deep Forest and Its Application to Automatic Detection of Cash-Out Fraud[J]. ACM Transactions on Intelligent Systems and Technology, 2019, 10(5): 1-19.

[43]王浩旻. 基于代价敏感和集成学习的网络借贷信用评价方法与应用[D]. 成都：电子科技大学，2020.

[44]Breiman L. Bagging Predictors[J]. Machine Learning, 1996, 24: 123-140.

[45]Schapire R E. The Strength of Weak Learnability[J]. Machine Learning, 1990, 5: 197-227.

[46]Finlay S. Multiple Classifier Architectures and Their Application to Credit Risk Assessment[J]. European Journal of Operational Research, 2011, 210(2), 368-378.

[47]沙靖岚. 基于LightGBM与XGBoost算法的P2P网络借贷违约预测模型的比较研究[D]. 辽宁: 东北财经大学, 2017.

[48]Papouskova M, Hajek P. Two-Stage Consumer Credit Risk Modelling Using Heterogeneous Ensemble Learning[J]. Decision Support Systems, 2019, 118, 33-45.

[49]周婧玥. 基于P2P借贷模式的互联网金融风险及其监管分析[J]. 商业经济, 2016(2): 81-82.

[50]刘洋，王会战. 互联网视角下我国民间借贷的风险评价与控制——以P2P平台为例[J]. 宏观经济研究，2017(3): 146-157.

[51]李晶，宋陆军. 我国P2P网络借贷平台违约问题及对策研究[J]. 会计之友, 2017(9): 79-83.

[52]Berger S C, Gleisner F. Emergence of Financial Intermediaries in Electronic Markets: the Case of Online P2P Lending[J]. BuR Business Research Journal, 2009, 2(1): 39-65.

[53]Xu C, Zhou Z, The Study of Internal Control and Over-Investment on Corporate Credit Risk[J]. Procedia Computer Science, 2016, 91: 109-113.

[54]Li L, Yang J, Xin Z, A Study of Credit Risk of Chinese Listed Companies: ZPP Versus KMV[J]. Applied Economics, 2016, 48(29): 2697-2710.

[55]Tsai S B, Li G, Wu C H, Zheng Y, Wang J. An Empirical Research on Evaluating Banks’ Credit Assessment of Corporate Customers[J]. SpringerPlus, 2016, 5: 2088.

[56]Kang M Y, Ausloos M. An Inverse Problem Study: Credit Risk Ratings as a Determinant of Corporate Governance and Capital Structure in Emerging Markets: Evidence from Chinese Listed Companies[J]. Economies, 2017, 5(4): 47.

[57]Han W, Wang P, Xu H, Choi J S. Evaluation of the Reverse Mortgage Option in Hong Kong[J]. Asian Economic Journal, 2017, 31(2): 187-210.

[58]Riggins F J, Weber D M. Information Asymmetries and Identification Bias in P2P Social Microlending[J]. Information Technology for Development, 2017, 23(1): 107-126.

[59]Dörr S, Raissi M, Weber A, Credit-Supply Shocks and Firm Productivity in Italy[J]. Journal of International Money and Finance Money Finance, 2018, 87: 155-171.

[60]Lin F, Xie S Y, Yang J P. Semi-Analytical Formula for Pricing Bilateral Counterparty Risk of CDS with Correlated Credit Risks[J]. Acta Mathematicae Applicatae Sinica, English Series, 2018, 34: 209-236.

[61]Stern C, Makinen M, Qian Z. FinTechs in China-with a Special Focus on Peer to Peer Lending[J]. Journal of Chinese Economic and Foreign Trade Studies, 2017, 10(3), 215-228.

[62]Galloway I. Peer-to-Peer Lending and Community Development Finance[J]. Community Investments, 2009, 21(3): 19-23.

[63]刘盼，李华康，孙国梓. 基于短时多源回归算法的P2P平台风险观测方法[J]. 计算机科学，2018, 45(5): 97-101.

[64]何浩然. 基于数据挖掘的P2P网络借贷信用风险评估研究[D]. 安徽：安徽财经大学，2017.

[65]王重仁，王雯，佘杰，凌晨. 融合深度神经网络的个人信用评估方法[J]. 计算机科学，2020, 46(10): 308-314.

[66]李淑锦，嵇晓佳. LGB-BAG在P2P网贷借款者信用风险评估中的应用[J]. 技术经济，2019, 38(11): 117-124.

[67]刘忻梅，唐俊，段翀. AUCRF算法在信用风险评价中的特征选择研究[J]. 计算机应用与软件，2018, 35(4): 293-295.

[68]Markowitz H M. The Utility of Wealth[J]. Journal of Political Economy, 1952, 60: 151-158.

[69]杨申. 步步高商业连锁股份有限公司投资管理研究[D]. 长沙：湖南大学，2017.

[70]宋鹏. 基于序化机理的稳健型股票价值投资研究[D]. 太原：山西大学，2012.

[71]褚中华. 中国DC型企业年金投资策略研究[D]. 北京：华北电力大学，2009.

[72]Ong C S, Huang J J, Tzeng G H. Building Credit Scoring Models Using Genetic Programming[J]. Expert Systems with Applications, 2005, 29(1): 41-47.

[73]胡晓丽，成力为. 国外商业银行信贷风险管理中国别风险的评估方法评介及启示[J]. 浙江金融，2012, (5): 50-52.

[74]Serrano-Cinca C, Gutiérrez-Nieto B, López-Palacios L. Determinants of Default in P2P Lending[J]. PloS ONE, 2015, 10(10): 1-22.

[75]Carmichael D. Modeling Default for Peer-to-Peer Loans[J]. Available at SSRN, 2014: 43.

[76]张亚京. 基于违约状态鉴别的农户小额贷款信用评级模型研究[D]. 大连：大连理工大学，2019.

[77]迟国泰，张亚京，石宝峰．基于Probit回归的小企业债信评级模型及实证[J]．管理科学学报, 2016, 19(6): 136-156.

[78]魏韡. 中国商业银行操作风险资本的测度——以中小商业银行为例[D]. 上海：复旦大学，2013.

[79]邹玉江. 基于机器学习的沪深300指数走势预测研究[D]. 济南：山东大学，2018.

[80]陈志达. 基于贷款明细数据的银行个贷信用评估实证研究[D]. 成都：电子科技大学，2018.

[81]Erdal H, Karahanoğlu İ. Bagging Ensemble Models for Bank Profitability: An Emprical Research on Turkish Development and Investment Banks[J]. Applied Soft Computing, 2016, 49: 861-867.

[82]Assouline D, Mohajeri N, Scartezzini J L. Large-Scale Rooftop Solar Photovoltaic Technical Potential Estimation Using Random Forests[J]. Applied Energy, 2018, 217: 189-211.

[83]He H, Zhang W, Zhang S. A Novel Ensemble Method for Credit Scoring: Adaption of Different Imbalance Ratios[J]. Expert Systems with Applications, 2018, 98: 105-117.

[84]Wolpert D H. Stacked Generalization[J]. Neural Networks, 1992, 5(2): 241-259.

[85]Fawcett T. An Introduction to ROC Analysis[J]. Pattern Recognition Letters, 2006, 27(8): 861-874.

[86]戴国强，肖海燕. 基金的业绩评估与组合管理[J]. 当代财经, 2003, (6): 38-40.

[87]姚志伟. 商业银行住房抵押贷款的利率和信用风险研究[D]. 武汉：武汉理工大学，2005.

[88]Provost F, Fawcett T. Data Science for Business: What You Need to Know About Data Mining and Data-Analytic Thinking[M]. Sebastopol, CA:O'Reilly Media, 2013.

[89]Guo Y, Zhou W, Luo C, Liu C, Xiong H. Instance-Based Credit Risk Assessment for Investment Decisions in P2P Lending. European Journal of Operational Research[J], 2016, 249(2): 417-426.

[90]Chi G, Ding S, Peng X. Data-Driven Robust Credit Portfolio Optimization for Investment Decisions in P2P Lending[J]. Mathematical Problems in Engineering, 2019, 2019: 1-10.

[91]田新民. 国际金融危机原因与对策思考[J]. 经济与管理研究, 2009，(10): 56-60.

[92]潘敏洁. 中国P2P网络借贷有限理性投资行为及决策辅助研究[D]. 上海：东华大学，2015.