摘要
本文针对我国互联网征信机构信用评分机制可能存在的缺陷,以"芝麻信用"为例,通过网络问卷调查方式,收集340位"芝麻信用"用户的线上信息作为研究样本集,在对样本集进行缺失性检查、逻辑性校验以及定量转化和排序等预处理工作后,使用多种相关性检验方法和半对数回归建模方法,研究用户线上信息对芝麻信用分具体的影响程度。在此基础上,结合相关理论和实际情况,分析了互联网征信评分机制过度依赖"淘气值"等信息、忽视收入和职业信息、基础信息全面性和真实性不足三大缺陷,并从如何改进"淘气值"、收入和职业信息依赖度、改善基础信息采集范围和质量以及如何尽快吸纳央行个人征信数据三个方面,对促进我国互联网征信行业发展提出了针对性的改进建议。
This paper aims to investigate the potential defects of credit scoring system used by IT companies in China with 'Zhima Credit' as an example. Firstly, we collect 340 Zhima Scores through online questionnaire survey and pre-treat the sample data by means of deletion test, logical verification, and data quantitative transformation and sorting. Secondly, we use a variety of correlation test methods and semi-logarithm regression models to find the influence of users' information on their 'Zhima Credit' score. Thirdly, taking into account relevant theories and the reality, we conclude that Internet Credit scoring mechanism has three main defects: excessive reliance on 'Taoqi Index', ignoring income and career information and the lack of comprehensiveness and authenticity of basic information. Finally, to promote the internet credit scoring industry development, we proposes suggestions on how to balance the reliance of 'Taoqi Index', income and occupation, how to improve the range and quality of basic information, and how to get central bank's database of personal credit as soon as possible.
出处
《金融监管研究》
2017年第9期48-65,共18页
Financial Regulation Research
基金
教育部人文社会科学研究青年基金项目<我国商业银行内部评级系统验证研究>(No.12YJC790116)
天津财经大学2017年青年教师自主选题预研资助计划的资助
关键词
互联网征信
个人信用评分
芝麻信用
大数据分析
Internet Credit Investigation
Personal Credit Scoring
Zhima Credit
Big Data Analysis