Research On Tax Compliance Risk Identification Based On Random Forest Algorithm

With the deepening of economic system reform in an all-round way,the tax environment has changed,and the game between tax authorities and taxpayers is serious.In September 2014,the state administration of taxation issued the "opinions on strengthening tax risk management",which proposed to take risk management as the guidance,take various data of tax declaration,collection and management as the support,and use information tools to strengthen the standardization of tax risk management.Tax compliance risk identification is the basis of tax risk management.With the comprehensive promotion of III golden tax project,the data provided by taxpayers to tax authorities are financial data and tax declaration data.Generally speaking,the production of financial data follows roughly the same accounting standards and reflects the production and operation status of enterprises through relatively common accounting methods.It is the main economic basic data for tax calculation.Using machine learning algorithms such as random forest to construct and train the risk identification model of tax compliance,mining the large data of tax revenue deeply,through the correlation analysis of the whole sample,eliminating the interference of human factors,predicting the corresponding tax amount,and comparing with the actual tax amount declared by enterprises,we can effectively identify the risk of tax compliance and rank its importance.At the same time,it assists tax staff in tax compliance risk management.At present,based on the basic big data such as taxpayer declaration and financial data,there arefew researches on tax compliance risk identification by machine learning.This paper,according to the theory of tax compliance risk and random forest algorithm,distinguished the difference between tax compliance risk and other tax-related risks,summarized tax compliance risk identification methods at home and abroad.And after introducing the case background of the application of random forest algorithm in tax compliance risk identification of B tax bureau,this paper expounded the problems existing in B tax bureau's tax compliance risk identification,such as the setting of early warning value depends on experience judgment and the difficulty in defining key indicators.According to the data mining process,the application of the model is analyzed from business understanding,data collection and pre-processing,modeling,model checking,result output and model evaluation.The quantitative measurement of tax compliance risk is realized,and the importance ranking is carried out.It is concluded that the output results of random forest algorithm should be treated objectively,data management should be strengthened,talent training system of big data should be established,and tax professional management should be implemented based on big data analysis.In addition to,it is hoped that this paper can provide some reference significance for tax authorities in China to identify tax compliance risk.