摘 要小微企业信用评分模型是有效提高小微授信业务高标准发展的重要研究课题，是金融机构风险管理技术的核心。本论文将基于数据挖掘的Logistic逻辑回归理论应用于构建小微企业信用评分模型问题的理论和应用研究，对于目前利率市场化加快、存贷息差收缩、互联网金融冲击下的小微企业的业务发展趋势具有重要的理论和现实意义。针对目前中国经济下行压力依然较大、发展不确定因素依然较多、结构性矛盾依然突出的情况下，研究着力服务好代表生机活力的小微企业有助于经济平稳发展，稳定抵御风险能力。与此同时，银行业面临信用风险、市场风险、流动性风险等市场的变革与创新也对银行业未来需要着力调整信贷结构，增加扶持小微企业贷款力度提出了更高的要求。本文以Logistic逻辑回归理论为指导，以某股份制商业银行二线城市分行的小微企业客户数据为支撑，针对如何有效构建信用评分模型问题，筛选有效样本数据，分析国内外金融机构的个人信用评分模型构建的参考因素，确定较为适合我国特色发展现状的小微企业信用评分模型体系，建立小微企业个人信用评分模型的评分方法。首先，论文分析了基于数据挖掘的Logistic逻辑回归理论，并基于该理论的基本原理建立小微授信审批评分模型，紧接着，论文深入研究模型的开发和实施，以及如何把评分模型引入到贷款审批环节。该论文详细介绍了模型开发前期数据的筛选，样本的选择以及样本的抽样，分析不同行业客户的行为特征，运用statal2.0软件平台，构建回归模型和分类技术，在此基础上，采用国内小微业务领先的民生银行小微客户案例，进行信用评分模型的实证分析，证明了小微企业授信评分模型的有效性，对于提升贷款审批效率及降低人为审批的主观性、道德性风险做了有效的数据验证。模型估计结果显示在测量的10个变量因素中影响评分结果前三位的依次是家庭个人净资产、企业上一年度的主营业务的销售收入和企业主从事行业经验长短，尤其以家庭个人净资产对个人信用评分有最为关键的影响，家庭个人净资产的增加能大幅度的降低个人信贷客户的违约概率，经营企业上一年度销售收入量的大小对于信贷违约测量结果是负相关，随着收入增多违约几率降低，企业主的从事行业年限增加能降低个人信贷客户的违约，提高客户的评分结果。另外其他解释变量对银行信贷违约也存在显著的影响。这表明通过运用逻辑回归Logistic方法建立起来的信用评分模型，具有较好的风险控制能力和客户分层划分，各项检验结果较为合理。本研究对国内商业银行建立小微企业经营贷款的信用评分模型具有参考价值，该技术运用有助于解决小微企业贷款成本高、风险概率大及信息不对称的突出问题。关键词：小微企业；经营贷款；信用评分模型；Logistic回归模型THE RESEARCH OF BANK CREDIT SCORING MODEL BASED ON DATA MINING FOR MICRO BUSINESSESABSTRACTSmall and micro enterprises (abbreviated as SMEs) Credit scoring model，which aimed at efficiently increasing the standards of small micro credit business, is an important research topic .In the same time it is the core technology in risk management for financial institutions.This paper will apply logistic regression theory based on data mining to theory and application research whcih help construct SMEss credit scoring model. In the face of acceleration of market-oriented interest rate reform，deposit and loan spreads shrinking，and the impact of Internet finance，SMEss credit scoring model has theoretical and practical significance for the business development of SME.As the Chinese downtown pressure on the economy is still large, the uncertain factors in current development are still more, structural contradictions are still prominent, good service for SMEs ,which represent economic vitality, contributes to the development of economic stability and the ability against risks. At the same time, faced with credit risk, market risk, liquidity risk and other market reform and innovation ,its critical to adjust the credit structure for banking industry and increase Loan limit for SMEs .This paper takes logistic regression theory as the guide, and uses a SMEss customer data from a branch of joint-stock commercial bank in second line city . Aiming to effectively constructing the credit scoring model, sample data was efficiently screened.Also the paper analyze the factor that domestic and foreign financial institution use to construct personal credit evaluation model , determining what is the more suitable SMEs credit scoring model system for the current China and establishing personal credit scoring method for SMEs.Firstly, the paper analyzes the logistic regression theory which based on data mining,and establishes SMEss credit model according to thistheory.Then,thethesisdeeply research the development and implementation of the model, and how the scoring model is introduced to the loan approval process. The paper details date screening in early stage of model development, the choice of samples and the sampling, and the analysis of behavior characteristics of customers in different industries.Using SPSS software platform, the logistic regression models and classification technology are built. On this basis, using SMEs customer case from MinSheng Bank and empirical analysis of credit scoring model, the validity of SMEs credit scoring model is proved,and provide data validation in improving the efficiency of examination and approval of loan and reduce the risk of artificial subjectivity.The results of model estimation show that among 10 variables, personal net assets, annual enterprises sales revenue from main business, and the length and experiences enterprise engaging in business are the more influential factors. Especially the personal net assetsis the most dominant factor and the increase inpersonal net assets can greatly reduce the probability of default. The amount of enterprises annual sales revenue as well as personal income is in negative correlation with default probability .the time length enterprises engaging in business can increase its credit and improve customer evaluation results. In addition, other explanatory variables are significantly influential in credit default.This suggests that credit scoring models set up by using logistic regression has the ability of customer stratification and better risk control, and the test results are reasonable. This research is of great referential value for domestic commercial banks to establish SMEs credit scoring model .Also the technology can help solve the problem of high cost loans for SMEs , high Risk probability,and Information asymme