基于Logistic模型的汽车金融公司个人贷款信用评分研究

发布时间：2018-05-28 14:59

本文选题：汽车金融公司 + 个人汽车贷款风险　；参考：《西南大学》2017年硕士论文

【摘要】：我国现正处于经济高速发展时期,汽车产业正是支撑经济增长的重要组成部分。而在整个汽车产业链中,汽车金融业的利润占到了近20%。汽车金融公司作为专业从事该行业的金融机构,随着利润的增长近几年内相继成立。中国个人汽车按揭贷款买车量占汽车销售总量的比例,即贷款渗透率在10年前尚不足5%,目前已经达到25%-35%,但距离很多发达国家70%以上的渗透率相比还有很大的差距。因此,我国汽车金融市场有着巨大的发展潜力,到2020年中国汽车金融的渗透率将达到50%,市场规模将达到2万亿元。但与国外成熟汽车金融市场相比,目前国内汽车金融市场还很混乱。由于存在信用体系建设不完善、个人收入信息不透明、地域广阔、人口流动性较大等客观原因,以及贷款审核效率不能满足市场需求、个人信用评级不完善等主观因素,造成了汽车金融公司违约风险的增加。因此,研究汽车金融公司如何通过个人信用评级来有效地控制违约风险就具有理论和现实意义。本文首先对国内外汽车金融的历史发展和现状进行了研究,并对国内汽车金融公司当前存在的问题及风险进行了阐述。然后对国内某大型汽车金融公司近三年个人汽车贷款客户的信息和还款记录进行了分析,研究如何通过建立信用评分模型来有效的对客户进行风险等级评估,从而提高审核效率和降低违约风险。研究过程中抽取了该公司近3年约25000名客户的资料,并通过问卷调查、借鉴行业先进经验等方式从基础信息、贷款信息及征信信息等筛选出了对个人汽车贷款风险有显著影响的8个变量,然后利用其中18592名客户进行Logistic回归建立个人信用评分模型,7970个客户用于验证个人信用评分模型区分能力。经分析检验表明:建立的评分模型的所有变量回归系数为负数,WALD检验P0.05,模型变量趋势与实际业务含义一致;方差膨胀系数VIF10,模型不存在多重共线性;K-S值为32.59,GINI系数为44.82,模型对好坏账户有较好的区分能力。最后根据评分模型对个人汽车贷款客户进行信用评级,根据其评级结果审核人员对客户实行差异化的审核,有效地提高了审核效率,还能较好地控制了个人汽车贷款风险。研究过程中还发现,个人数据资料的真实性、完整性是保证评分模型可靠的关键。同时,本文还研究提出了提高汽车金融公司风险防范能力的措施和办法:加强内部培训,提高审核人员综合素质和业务技能;根据等级评分模型统一审核要求;建立审核人员的资格认证,建立淘汰制度;针对不同评级客户实行差异化审核政策和建立不同的金融产品;建立完善贷后管理,建立风险共担的金融风险体制,明确单个客户风险监控主责任人,建立责任人负责制度和重点客户管理制度。
[Abstract]:China is now in the period of rapid economic development, automobile industry is an important part of supporting economic growth. But in the whole automobile industry chain, the automobile finance industry profit accounted for nearly 20%. As a financial institution specialized in this industry, auto finance company has been established in recent years with the increase of profit. China's personal car mortgage loans as a proportion of total car sales, that is, the loan penetration rate of less than 5% 10 years ago, has now reached 25-35, but there is still a big gap between the penetration rate of more than 70 percent in many developed countries. Therefore, China's auto financial market has great potential for development. By 2020, the penetration rate of China's auto finance will reach 50%, and the market scale will reach 2 trillion yuan. But compared with the foreign mature automobile finance market, the domestic automobile finance market is still very chaotic. Because of the imperfection of credit system construction, the opaque personal income information, the vast area, the large population mobility, and other subjective factors, such as the loan audit efficiency can not meet the market demand, the personal credit rating is not perfect and so on. Caused the auto financing company default risk increase. Therefore, it is of theoretical and practical significance to study how auto financing companies can effectively control default risk through personal credit rating. In this paper, the historical development and present situation of automobile finance at home and abroad are studied, and the existing problems and risks of domestic auto finance companies are expounded. Then it analyzes the information and repayment records of a large auto financing company in China in the past three years, and studies how to effectively evaluate the risk rating of customers by establishing a credit rating model. In order to improve the efficiency of audit and reduce the risk of default. In the course of the research, the data of about 25000 clients of the company in the past three years were extracted, and the basic information was obtained from the basic information by means of questionnaire survey, advanced experience of the industry and so on. The loan information and credit information have screened out 8 variables that have significant influence on the risk of personal automobile loan. Then 18592 of them were used for Logistic regression to establish personal credit rating model, and 7970 customers were used to verify the distinguishing ability of personal credit rating model. The analysis and test show that the regression coefficient of all variables in the established scoring model is negative and WALD test (P0.05), and the trend of the model variable is consistent with the actual business meaning. The coefficient of variance expansion is VIF10, and the K-S value of multiplex collinearity is 32.59 and the coefficient of Gini is 44.82. The model has a good ability to distinguish between good and bad accounts. Finally, according to the rating model, the credit rating of individual car loan customers is carried out. According to the rating results, the auditor carries out the differentiated audit to the customers, which effectively improves the audit efficiency and controls the risk of personal automobile loans. It is also found that the authenticity and integrity of personal data is the key to ensure the reliability of the scoring model. At the same time, this paper also studies and puts forward the measures and methods to improve the risk prevention ability of auto finance companies: strengthening internal training, improving the comprehensive quality and professional skills of auditors, unifying the audit requirements according to the rating model; Establish qualification certification of auditors, establish elimination system; implement differentiated audit policy and establish different financial products for different rating clients; establish and perfect post-loan management, establish a risk-sharing financial risk system, Identify the main responsible person for individual customer risk monitoring, establish the responsible person responsibility system and key customer management system.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.4

【参考文献】