贝叶斯多重填补法在食品企业信用评级的应用研究

发布时间：2019-03-12 11:43

【摘要】：2008年美国的次贷危机演变成一场全球金融危机，世界各国经济都遭受到严重的打击。与此同时，中国爆发了“三鹿奶粉”事件，造成中国民族乳业品牌的一次集体大溃败,对中国乳业国际声誉的产生了致命影响。食品企业这些问题严重影响了食品行业信用情况，一家企业受到食品安全问题的影响后，企业信用可以说是降到谷底，食品企业如何彻底应对这一危机影响，成为企业信用建设的重点问题。本文以信用评级的初始步骤一一数据预处理为出发点，研究有缺失数据的数据集的填补问题。过去的企业信用评分研究多着眼于模型的设计，而忽视了对不完整数据预处理的研究。事实上，比较成熟的建模方法对其处理的数据集都有一定的要求，如数据完整性好、冗余性低、具有代表性等。数据缺失是信用数据中普遍存在的问题，对它的处理已成为信用评分研究中的关键问题。在我们做信用评级的研究过程中，难免会遇到这样那样的数据缺失问题，那么如何处理这一问题就成为了影响后续研究的关键问题。本文根据前人的经验和研究结果，运用基于贝叶斯理论的多重填补法来对有缺失数据的数据集进行填补，然后在对填补后的数据进行统计分析，最后将数据代入到信用评级模型中，来得到食品企业的信用状况。本文运用SAS统计软件对缺失数据集进行填补处理，得出的结果给出了5组填补值，分别计算得出29个食品企业的信用评分的排名情况，本文认为，，这种方法在一定程度上克服了单一填补法在数据填补上的局限性，很好的处理了缺失数据的不确定性，为后续研究奠定了坚实的数据基础。本文通过填补方法的介绍，食品企业信用现状的分析，阐述了数据填补对于评级研究的意义，最后用实证研究给出了填补后的研究结果，与传统的方法相比，本文的方法具有示范性的作用，能更好的运用到食品企业信用评级中去，甚至对所有样本的数据缺失都具有可操作性。
[Abstract]:In 2008, the subprime mortgage crisis in the United States became a global financial crisis, and the world economy suffered a serious blow. At the same time, the "Sanlu milk powder" incident broke out in China, resulting in a collective debacle of the Chinese national dairy brand, which had a fatal impact on the international reputation of the Chinese dairy industry. These problems have seriously affected the credit situation of the food industry. After an enterprise is affected by the food safety problem, the enterprise credit can be said to have dropped to the bottom. How do the food enterprises deal with the impact of this crisis thoroughly? Become the key issue of enterprise credit construction. This paper takes the initial step of credit rating-data preprocessing as the starting point, and studies the filling problem of data sets with missing data. In the past, the research of enterprise credit score mostly focused on the design of model, but ignored the research of incomplete data preprocessing. In fact, the mature modeling methods have certain requirements for the data set, such as good data integrity, low redundancy, representative and so on. The lack of data is a common problem in credit data, and the processing of it has become a key issue in the research of credit rating. In the course of our credit rating research, we will inevitably encounter the problem of data loss, so how to deal with this problem has become a key issue affecting the follow-up research. According to the previous experience and research results, this paper uses the multi-filling method based on Bayesian theory to fill the data set with missing data, and then carries on the statistical analysis to the data after filling. Finally, the data are added to the credit rating model to get the credit status of food enterprises. In this paper, SAS statistical software is used to fill the missing data set, the result gives five groups of filling values, and the credit rating of 29 food enterprises is calculated respectively. To some extent, this method overcomes the limitation of single filling method in data filling, deals with the uncertainty of missing data well, and lays a solid data foundation for further research. Through the introduction of the filling method and the analysis of the credit status of food enterprises, this paper expounds the significance of data filling for rating research, and finally gives the results of the filled research by empirical research, compared with the traditional method. The method presented in this paper can be applied to the credit rating of food enterprises better, and even can be operated on the lack of data of all samples.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.4;F426.82

【参考文献】