Predicting Tax Evasion of Legal Taxpayers with an Emphasis on Economic Components, Taxpayers and Auditors Characteristics;Relying on Artificial Intelligence

Governments need stable and reliable financial resources to carry out their public duties, and taxes have long been one of the most important sources of funding for governments to carry out their duties. Preventing or reducing the amount of tax evasion has been one of the important concerns of governments in the economic field in recent years. In this research, the category of tax evasion using artificial intelligence and focusing on a set of 57 financial and non-financial indicators at the macro-economic level, taxpayers and tax auditors, in a sample including 978 legal taxpayers' files of the General Directorate of Tax Affairs of Mazandaran. It has been examined for the years 2011 to 2018. In this research, in order to extract effective features, sine cosine and gray wolf optimization algorithms were used, and decision tree and artificial neural network algorithms were used to model tax evasion and test features. Based on the results, the sine-cosine optimization algorithm along with the decision tree predictor has a lower error value than other models and provides a more accurate model for predicting tax evasion.