Evaluating The Effectiveness of Decision Tree Algorithm and Linear Multivariate Regression in Predicting Tax Evasion of Legal Taxpayers in Mazandaran Province

Tax revenues are one of the most important sources of government income and provide a major part of its expenses. The main goal of this research is which of the decision tree algorithm and linear multivariate regression methods provides a better prediction of tax evasion of legal taxpayers. Based on the theoretical foundations and background studies of a set of variables including 57 financial and non-financial indicators at three macroeconomic levels, taxpayers and tax auditors, in a sample consisting of 964 cases of legal entities at the level of the Mazandaran General Administration of Tax Affairs for the years 2012 to 2019 with The use of Python and Stata software has been investigated. At first, the sine-cosine identification algorithm was used to select the influencing variables. The results of the data analysis showed that the variables at the level of taxpayers and tax auditors are more effective in predicting tax evasion. Also, the findings indicate that the predictive power of the decision tree algorithm is higher than the linear multivariate regression