【作者】 张毅；

【导师】 刘立新；

【作者基本信息】 对外经济贸易大学 ， 统计学， 2022， 硕士

【摘要】 伴随国内金融市场开放程度加快,利率市场化逐渐向纵深推进,商业银行的市场竞争越发激烈,尤其在当前金融科技大力发展的时代,商业银行面临更大的机遇与挑战,如何利用好数据,发挥数据资产价值,实现零售业务的数字化转型,是商业银行提升核心竞争力的关键所在。商业银行发展零售金融业务的核心是加强个人客户价值管理,力求客户价值最大化。在实际经营中,就是要求银行对个人客户实现更为有效的营销、风险管理和综合价值评估,实现传统业务运营模式向更高效、客观、智能的“数据+模型”驱动的运营模式转变。借助“数据+模型”驱动的方式,降低银行相关工作人员在客户营销、风险评估、贷后催收等业务环节中因信息不对称、经验不丰富、流程不规范等因素导致的决策、操作、道德等风险,帮助客户极大地实现空间的自由和时间的节约,同时大大节省银行业务运营成本,促进零售金融业务快速、高质量发展,助推传统商业银行实现零售金融业务的数字化转型。本文首先介绍了商业银行零售信贷业务现状及面临的问题。其次,针对商业银行零售信贷业务面临的问题提出了数据模型的解决方案,对商业银行零售信贷业务在贷前申请、贷中监测预警、贷后催收等关键业务场景中常用的数据模型进行介绍。在贷款申请进件阶段,由于反欺诈客户特征—般隐藏较深,采用区分度较好的XGBoost机器学习算法构建反欺诈模型;在贷前审批、贷中监测、贷后催收各个阶段场景下,由于评分卡模型具有简单直观、解释性强、部署简便、监控迭代方便等优点,因此采用评分卡模型做为各个环节的评估模型。再次,结合实际案例介绍商业银行零售信贷业务贷前、贷中、贷后评估模型的构建过程,通过实际案例对商业银行零售信贷业务全流程评估模型进行实证分析,探索商业银行传统零售信贷业务数字化转型发展的有效路径。案例中对客户在商业银行内、外部的数据进行分析,内部数据包括客户交易数据、资产负债相关账户数据、风险偏好等数据,外部数据包括人行征信数据,以及外部第三方提供的客户黑名单、多头借贷、设备、运营商、司法等数据。通过对该些内、外部数据的处理分析,以及按照模型构建过程分步骤筛选处理,最终得到影响模型输出效果的关键模型变量,建立相应的数据模型,并通过数据模型的实际运用提升了案例银行在零售信贷业务领域整体的风险防控水平;同时,案例实现的线上风控模式,大力提升了案例银行在零售信贷业务方面的数字化水平。最后,对商业银行零售信贷业务全流程评估模型的研究成果进行总结,并展望该研究成果对商业银行零售信贷业务的适用性。更多还原

【Abstract】 Openness is accelerated,with the domestic financial market gradually to promote interest rate liberalization,the commercial bank market competition increasingly fierce,especially in the current financial developing era of science and technology,commercial Banks face more opportunities and challenges,how to make good use of the data,to display the data value assets,the realization of digital transformation of retail business,It is the key to enhance the core competitiveness of commercial banks.The core of commercial banks’ development of retail financial business is to strengthen individual customer value management and strive to maximize customer value.In practical operation,banks are required to realize more effective marketing,risk management and comprehensive value assessment for individual customers,so as to realize the transformation of traditional business operation mode into a more efficient,objective and intelligent "data+model" driven operation mode.+ with the help of a "data model" drive way.reduce the bank related personnel after the customer marketing,risk assessment,loan collection in the business such as link because of information asymmetry,imperfect experience rich,the process caused by such factors as decision-making,operation,such as moral risk,help customers to greatly to achieve freedom and time saving of the space,save the cost of business operations at the same time,It will promote the rapid and high-quality development of retail financial business and help traditional commercial banks realize the digital transformation of retail financial business.This paper first introduces the current situation and problems of retail credit business of commercial banks.Secondly,in view of the problems faced by the retail credit business of commercial banks,the solution of data model is proposed,and the data model commonly used in the key business scenarios of commercial banks’ retail credit business,such as pre-loan application,in-loan monitoring and early warning,and post-loan collection,is introduced.In the process of loan application,XGBoost machine learning algorithm with good discrimination was used to construct the antifraud model because the characteristics of anti-fraud customers were generally hidden deeply.In the scenarios of various stages of pre-loan approval,in-loan monitoring and post-loan collection,the scorecard model is used as the evaluation model of each link because of its advantages such as simple and intuitive,strong interpretation,easy deployment and convenient monitoring and iteration.Thirdly,this paper introduces the construction process of pre-loan,on-loan and post-loan evaluation models of retail credit business of commercial banks based on actual cases.Through the empirical analysis of the whole process evaluation model of retail credit business of commercial banks,this paper explores the effective path of digital transformation and development of traditional retail credit business of commercial banks.Cases to customers in the commercial bank’s internal and external data analysis,the internal data including customer transaction data,account data related to assets and liabilities,risk preference data,such as external data including the pedestrian reference data,as well as the external third party customer list,the bulls borrowing,equipment,operators,judicial and other data.Through processing the some internal and external data,analysis,and according to the model building process step screening process,the final key model parameters that influence the effect of model output variable and the corresponding data model,and through the practical application of the data model to improve the case bank overall level of risk prevention and control in the field of retail credit business;At the same time,the online risk control mode realized by the case greatly improves the digital level of the case bank in retail credit business.Finally,the paper summarizes the research results of the whole process evaluation model of commercial banks’ retail credit business,and prospects the applicability of the research results to commercial banks’ retail credit business.更多还原