Tax Audit Selection by Using of Data Mining Algorithms

Since the direct taxes law was approved in 2014, and its article 97 was amended, the State tax affairs organization has been required to accept tax returns from individuals whose financial year begins on 29/07/2018, and to select and audit only a few of those returns, based on risk indicators. Using data mining methods, it is possible to determine high-risk taxpayers based on their information. In this way, high-risk taxpayers can be identified. During this study, tax returns information for legal entities from 2014 to 2016 was used in order to assess the level of risk.. Finally, the success of the methods has been evaluated. The algorithms used are vector machine classification methods, neural network support, decision tree and nearest neighbor. The results of the research confirm that the neural network algorithm is introduced as the best algorithm for estimating the risk of the statement.