摘要
纳税评估是指税务机关根据纳税人的申报资料及其他征管信息,按照一定的程序,运用一定的手段和方法,进行审核、比对、分析、核查,对纳税人一定时期内申报纳税的真实性、准确性进行综合评估的工作。纳税评估工作在整个税收管理体系中处于重要位置。而作为整个纳税评估工作的源头,确定纳税评估对象这一步骤意义重大。建立有效的纳税评估模型,能够为纳税评估对象的确定提供较为客观的依据,使这一工作从定性判断的阶段发展到定量判断阶段,并将对税务部门的纳税评估日常工作产生深远而实质性的影响。
贝叶斯统计理论作为当前两大统计学派之一,在西方国家有着广泛的应用。国内外的众多文献资料表明,贝叶斯统计理论作为一种基础研究工具已经广泛应用于各个学科,在经济预测、垃圾邮件识别、可靠性推断、财务预警、信用评估、客户细分等具体领域都有着深入的应用。与其他模型相比,基于贝叶斯方法的模型在稳定性与准确性方面有其优势,因此贝叶斯方法得到了众多领域专家的认同。其中朴素贝叶斯分类技术在对象属性分类上已经有着众多成熟应用。
纳税评估对象的确定问题实质上也是一种纳税人的分类问题,本文的研究核心就是将朴素贝叶斯分类技术应用到税务行业中,形成基于这一技术的纳税评估模型,从而为纳税评估对象的确定提供一个定量判断的手段。
本文以纳税评估模型作为研究对象,首先考察了当前国内外在纳税评估模型领域的研究现状;在分析纳税评估模型实质及概述贝叶斯分类技术的基础上,提出建立基于朴素贝叶斯分类技术的纳税评估模型的思路;然后依据这种思路设计并实现一种基于朴素贝叶斯分类技术的纳税评估模型。
通过使用实证数据测试,本文所研究的纳税评估模型在纳税评估业务最关心的“税收不遵从”分类预测方面准确率与召回率比较理想,在税收业务中辅助筛选异常纳税人方面具有实际应用价值。并且本文所研究的纳税评估模型计算复杂度较低、模型简洁,容易实现并应用到实际业务中。本文的研究工作包括以下部分:
一、分析纳税评估及纳税评估模型研究的背景、意义,考察国内外在此领域的研究现状。
二、分析纳税评估模型的实质,并对贝叶斯分类技术相关理论基础作必要说明。
三、结合纳税评估模型这一具体领域,提出朴素贝叶斯分类技术在这一领域的相关描述。
四、对纳税评估模型涉及到的具体业务及相关数据准备情况作详细说明。
五、设计基于朴素贝叶斯分类技术的纳税评估模型并实现相应的软件功能。
六、使用具体的业务数据对基于朴素贝叶斯分类技术的纳税评估模型及软件进行评估、验证,提出这一研究方法的局限性与未来应用的拓展方向。
Tax auditing is the process that tax bureau analyzed if taxpayer had finished the obligation of tax trustily, by rule and line. The analysis is based on the information taxpayer submitted and the information fetched by tax bureau through other means. Tax auditing is a very important part of the whole tax management. The foundation of tax auditing is to find out which taxpayer is the one needed more attention. If there is an efficient analyzing model as assistant, the process of tax auditing will be more objective and easy.
As one of the two major statistics theory, Bayesian statistics theory had been applied widely in the west world. Massive information proved that Bayesian statistics theory can be used to forecast economy development, recognize the spam email, conclude the reliability, make the early warning of company finance, analyze the credit, classify the customer, and so on. Bayesian method has the advantage over other methods by its stability and veracity. So there are lots of domain experts would like to make research with Bayesian method as a basic tool. Also there are lots of applications on classifying field with na?ve Bayesian classifying method.
To find out which taxpayer is the one needed more attention, it is a kind of classifying. The core of this study is to apply the naive Bayesian method on taxation management, further more to construct a tax auditing model based on naive Bayesian classifying method.
Tax auditing model is the study object of this paper. At first, I study the actuality of tax auditing model in the in the world wide. By analyzing the essence of tax auditing and summarizing the Bayesian classifying method, I advanced the idea to construct a tax auditing model based on naive Bayesian classifying method. Further more, I designed and implemented this model using the IT technology.
The index of Precision and Recall of this tax auditing model is ideal on the test by actual data. It is valuable on the taxation management. In addition, this tax auditing model is simple and easy to use in the actual operation.
The following is what I study through this paper:
1. Describe the background and meaning of tax auditing and tax auditing model. Studying the actuality of tax auditing model in the in the world wide.
2. Analyze the essence of tax auditing and summarizing the basis of Bayesian classifying method.
3. Define the domain of tax auditing by using na?ve Bayesian classifying method.
4. Define the details of actual tax auditing and the interrelated data.
5. Design and implement the tax auditing model based on the na?ve Bayesian classifying method by using IT technology.
6. Evaluate the tax auditing model and the corresponding software by using the actual operation data. Find out the limitation of this method and the farther developing direction.
引文
[1]茆诗松.贝叶斯统计[M].北京:中国统计出版社,1999.9
[2]王双成.贝叶斯网络学习、推理与应用[M].上海:立信会计出版社,2010.2
[3] OECD, Information Note: Compliance Risk Management Audit Case Selection Systems: Case Study, 2004.10, 14—45
[4] Helen V. Tauchen; Ann Dryden Witte;Kurt J.Beron.Tax compliance: An investigation using individual taxpayer compliance measurement program (TCMP) data.Journal of Quantitative Criminology.1993-6
[5] Nipoli Kamdar .Information reporting and tax compliance: An investigation using individual TCMP data.Atlantic Economic Journal.1995-12
[6] B.Tran-Nam, C.Evans, M.Walpole and K. Ritchie,"Tax Compliance Costs :Research Methodology and Empirical Evidence from Australia," National Tax Journal, 53,2,229-52.
[7]周伍阳,杨招军.“纳税评估”理论基础及其指标体系研究[J].石家庄经济学院学报.2005(2)
[8]陈继阳.纳税评估体系研究[D].东北财经大学.2006.12
[9]彭十一,周伍阳.Logit模型在纳税评估中的应用[J].统计与决策.2008(5)
[10]徐戎,王文杰,周四新.神经网络与领域知识结合的纳税评估预警模型[J].电子科技大学学报.2009.38(1)
[11]倪涛.税务数据仓库及其基于C4.5挖掘算法的纳税评估模型研究[D].国防科学技术大学研究生院.2006.5
[12]蔡伟鸿,郭陈熹.遗传算法优化BP神经网络在纳税评估中的应用[J].汕头大学学(自然科学版).2008.23(2)
[13]张凤娜.浅议构建税务审计的财务分析模型[J].现代经济信息.2009(13)
[14]石鑫.决策树分类算法的研究及其在纳税评估中的应用[D].中国海洋大学.2004.4
[15]王锐.税收不遵从的识别研究[学位论文].浙江大学.2003.04
[16]杨得前.税收遵从的理论研究及其在税收管理中的应用[D].上海理工大学2006.6
[17]金鹏,蔡淑琴.纳税评估信息系统对税收遵从的影响模型[J].情报杂志.2008.12
[18]朱慧明.贝叶斯多元统计推断理论.科学出版社.2006.9
[19]朱慧明.现代贝叶斯统计理论的基本观点与研究现状[J].江苏统计.2003.01.12-13
[20]岳金凤.贝叶斯方法在保险精算中的应用综述[D].吉林大学.2009.4
[21]朱慧明,陈骏武,马奔.基于贝叶斯网络学习模型的客户关系管理研究.统计与决策.2006.4
[22]郭雨松.一种启发式贝叶斯分类算法及其在铁路货运客户细分中的应用研究[D].北京交通大学.2008.6-10
[23]李旭升.贝叶斯网络分类模型研究及其在信用评估中的应用[D].西南交通大学.2006.6
[24]李旭升,郭耀煌.一种新颖混合贝叶斯分类模型研究[J].计算机科学。2006.33(9)
[25]康庄,余元全.基于贝叶斯分类器的纳税评估模型研究[J].经济问题.2009(6)
[26]汪洋.基于工作流的纳税评估建模研究和系统实现[D].中南大学.2006-06-30
[27]王翠霞.基于贝叶斯网络方法的上市公司财务预警模型[D].沈阳工业大学2006.3
[28]谭光荣.选择纳税评估指标的局限性及应对措施[J].税务研究.2007(02)
[29]王志铭.运用税负率进行纳税评估应注意的问题[J].商业会计.2008(08)
[30]王海勇,金菁.纳税评估制度的国际比较与借鉴[J].地方财政研究.2008(02)
[31]钟原.税收遵从风险管理探讨[D].首都经济贸易大学, 2007
[32]王久铱.基于聚类的税务稽查选案方法及其系统的研究[D].大连理工大学.2007
[33] Donna D.Bobek; Robin W.Roberts; John T. Sweeney.The Social Norms of Tax Compliance: Evidence from Australia, Singapore, and the United States.Journal of Business Ethics.2007-8