Study On The Application Of Rough Sets And Support Vector Machines In Taxation Evaluation

Taxation evaluation is a new project in domestic tax administration, but academic research is insufficient. Mainly, there are three main reasons for this: Firstly, someone emphasizes tax is an one-way relationship between rights and obligations, that tax collection is national rights and a taxpayer only fulfills the obligation to pay taxes, that the taxation is mandatory, fixed and gratis. Secondly, there is nothing to promote and enhance the administrative efficiency. The only object has been ensured for tax management that completes the designated budgetary revenue. Local financial management, that is the lack of supervision of the Budget, often determines tax revenue based on expenditure. Lastly, the government has no more released public information. The acquisition of research data is very difficult. As a result, from the origin and nature of the tax, this dissertation demonstrates that taxation evaluation is based on tax compliance theory. Then, the indicators are reduced by using the rough sets. Finally, the taxation evaluation model is established by support vector machines (SVM).The basis of the tax compliance theory is social contract that includes importantly exchange and needs of the public theory, which stresses taxation is contracted, certain and exchanged. Thus in essence, tax collection, which must ask permission, is actually self-management activities for the taxpayer. Based on this, tax compliance theory has been developing. The research of taxpayer compliance and taxpayer noncompliance phenomenon concerns two problems: discussing types of the taxpayer compliance and noncompliance; studying the compliance costs. Taxation evaluation is to reduce compliance costs for the taxpayer of compliance. The taxpayer has been selected for tax noncompliance. The tax authorities will improve management efficiency and reduce management costs, at the same time the taxpayer of compliance will reduce costs of compliance.In this dissertation, the Rough sets theory has been screened the indicators of taxation evaluation, which is based on previous studies about the "Structure Model" and the Principal Component Analysis screening indicators. The indicators of accounting, which reflect the tax compliance of taxpayer, are reduced from 19 to 5. The five indicators can be used as the indicators of taxation evaluation for they reflect the four conditions of taxpayer. Moreover, during the optimization of model parameters, grid-searching algorithm is used to study the relationship between parameters and prediction accuracy of SVM, and the relationship among themselves. The parameter group with the best prediction accuracy for SVM is selected to set up the final model structure of SVM.This dissertation also discusses the discrete method and chooses the SOM neural network for discrete data processing. According to the 76 taxpayers accounting indicators of transport industry, the taxation evaluation model has been established. The application of the model shows that the Rough sets theory on the indicators reduction of taxation evaluation are better capacity and the SVM method provides excellent modeling and great generalization abilities when applied to judge if the taxpayers paid taxes correctly. Therefore, this study can be directly applied to the tax authorities to establish taxation evaluation model and promote tax management.