（毕业论文）基于Logistic回归的个人信用评估模型及其改进--针对消费金融企业的个人信用评估.docVIP

摘 要 随着经济的日益发展，人们的消费日趋多元化，消费信贷应运而生，各大商业银行与消费金融业均致力于消费信贷产品的开发。但当消费信贷产品面向个体客户时，其中存在的违约风险是不可忽略的，因此消费金融行业对个体客户的信用评估尤为重要。我国信贷业务开展时间不长，还有很多不完善的地方，如信用制度不完善，信用风险的难测性等。此外，对信贷风险的管理也没有形成完整独立的体系，消费信贷风险逐渐暴露，成为制约信贷发展的一大障碍。因此本文针对信贷者的居住稳定性，就业稳定性，收入状况等方面的信息来判断信贷者的信用状况，从而估测信贷的违约的风险。 由于国内个人基本信息的保密性，我们采用加州大学欧文分校(UCI)提供的机器学习公开数据集中的德国信用数据集进行实证研究。 本文在充分理解影响个人消费信贷信用的各种指标的基础上（收入、年龄、受教育情况、就业稳定情况等属于基本参考指标），发散思维，引入两项新的指标----婚姻状况与应抚养人数。就已经考虑到的指标，对收集到的各项指标数据，根据各项指标的内在特点，采用合适的量化方法，对数据进行量化分析，对处理过的数据做主成分分析，提取其中影响信用评分的主成分，将得到的主成分带入logistics回归模型，利用极大似然法确定参数，对最终确定的参数进行验证后，分析模型的显著性、拟合程度及不足，得到信用评分与违约情况的联系最后建立信用评估系统。文章中给出了本文的两大创新点——新指标的引入与应对数据缺失的办法。 关键词：信用评估 数据量化 logistic模型 新指标运用 Abstract： With the development of economy, peoples consumption has become increasingly diversified, loan for consumption arises at the historical moment,. Many big commercial banks and consumer finance are committed to the development of credit products, But when consumer credit products increase to satisfy the needs of individual customers, there exist risks of default which should not be ignored. So it is particularly important to do credit assessment for individual customers. Because the history of our country’s credit businesses are not long, and there are many drawbacks, such as incomplete credit system, credit risk and capricious .The credit risk management does not form a complete independent system, so the gradually exposed consumer credit risks become the obstacle restricting the development of the credit system. So this article is in view of the stability of people lives, stability of employment, income, etc, to judge the credit information of credit conditions, so as to estimate the credit risk of default. Due to the confidentiality of the basic information of the individuals in China, we use the university of California, (UCI) providing public data sets of machine learning German credit data set for empirical research. Based on fully understand of the impact of pe