فرار مالیاتی در واردات ایران، رویکرد مدل ترکیبی شبکه عصبی مصنوعی و الگوریتم تبرید تدریجی

Import tax is one of the government revenue sources that some of its portion is not accessible to government due to tax evasion. In this study, the factors affecting tax evasion in import, have been identified by using the combinatorial model of artificial neural network and simulated annealing algorithms that is capable to analyze the nonlinear systems. For this purpose, four explanatory variables representing tax evasion in the Iran’s economy include tax burden on imports, the size of governments, tax payers’ real income and trade were considered in specified final model. The results from optimization of tax evasion using simulated annealingalgorithms indicate that the optimum of the burden tax on imports, government size, economic openness and real income per capita are 6.18 percent, 13.2, 6.69 Million Rials, 1.29 percent, respectively; Also, the minimum tax evasion in the period under analysis amounts to 21.48 percent.