数据挖掘在税收分析中的应用研究

发布时间：2018-06-01 00:00

本文选题：数据挖掘 + 税收分析　；参考：《财政部财政科学研究所》2014年硕士论文

【摘要】：随着人工智能和现代信息技术的不断发展,人类社会迈入了信息化时代,大量的信息充斥于各行各业。通过税收信息化系统的建设和推广应用,我国税务部门掌握了庞杂的能反应国民经济运行情况的原始税务数据。对这些数据进行分析,发现其所蕴含的有价值的潜在信息,有助于政府部门对社会主义市场经济进行宏观指导。我国传统的税收分析工作过于简单,仅限于一些定性的描述和同比数据分析,税收分析报告的内容不能满足各界人士的使用需求。国家税收征管系统中的大量数据没有得到有效的挖掘,不能传递出有用的信息,造成了信息资源的浪费。数据挖掘技术的发展为解决这种困境提供了技术方法。本文在学习数据挖掘理论和常用算法的基础上,结合我国现行税收分析工作的主要内容,提出了可以利用回归预测、K-均值聚类算法和因子分析对传统的税收分析报告做出一定的改进,并在之后的实证分析中验证了方法的可行性。首先利用回归分析对税收收入总量进行曲线拟合,发现我国税收收入总量的增长呈现出指数曲线的趋势；其次利用回归分析对税收收入总量和国内生产总值进行对数曲线拟合,回归曲线的自变量系数就是税收弹性的估计值；再利用K-均值聚类算法对我国省级行政单位进行聚类分析,分类依据是省级单位分税种税收收入,分类结果将我国省级行政单位划分了三大类,且经济意义较为显著；然后利用因子分析对上市公司的财务指标进行降维处理,目的是利用因子得分对企业所得税税源进行监控,结论是因子得分位于中间段的企业,财务报表被操纵导致税源不稳定的可能性比较大；最后利用同比增长平均值和对数回归曲线对税收收入进行预测,相较而言,指数曲线的预测效果较好。我国的税收信息化建设已取得一定成就,将数据挖掘技术引入到税收分析中有助于进一步提升税收管理工作的质量和效率。
[Abstract]:With the continuous development of artificial intelligence and modern information technology, human society has entered the information age, a large number of information flooded in various industries. Through the construction and popularization of the tax information system, the tax authorities of our country have grasped a lot of original tax data which can reflect the operation of the national economy. Through the analysis of these data, the valuable potential information contained in the data is found, which is helpful for government departments to guide the socialist market economy macroscopically. The traditional tax analysis work in our country is too simple, limited to some qualitative description and comparative data analysis, and the contents of tax analysis report can not meet the needs of people from all walks of life. A large number of data in the national tax collection and management system have not been effectively mined and can not transmit useful information, resulting in a waste of information resources. The development of data mining technology provides a technical method to solve this dilemma. On the basis of studying the theory of data mining and common algorithms, this paper combines the main contents of the current tax analysis work in our country. It is proposed that the traditional tax analysis report can be improved by using regression prediction K-means clustering algorithm and factor analysis, and the feasibility of the method is verified in the later empirical analysis. First, the regression analysis is used to fit the total tax revenue, and it is found that the growth of the total tax revenue in China shows an exponential curve trend; secondly, the logarithmic curve fitting of the total tax revenue and GDP is carried out by using the regression analysis. The independent variable coefficient of the regression curve is the estimated value of tax elasticity, and then using the K-means clustering algorithm to cluster the provincial administrative units in China, the classification is based on the tax revenue of the provincial units. The classification result divides the provincial administrative units of our country into three categories, and the economic significance is significant. Then, the factor analysis is used to reduce the dimension of the financial indexes of the listed companies, in order to monitor the enterprise income tax sources by using the factor score. The conclusion is that if the factor score is in the middle, the possibility of the financial statements being manipulated will lead to the instability of the tax source. Finally, the average growth and logarithmic regression curve are used to forecast the tax revenue, compared with the average value of the annual growth and the logarithmic regression curve. The prediction effect of exponential curve is better. Some achievements have been made in the construction of tax information in our country. The introduction of data mining technology into tax analysis will help to further improve the quality and efficiency of tax management.
【学位授予单位】：财政部财政科学研究所
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;F812.42

【参考文献】