Presenting a model for predicting tax evasion guilds based on Data mining techniques

In this research, considering the importance of the topic and deficiency in previous researches, amodel for predicting tax evasion of guilds based on data mining technique is presented. Theanalyzed data includes the review of 5600 tax files of all guilds holding tax codes in Qazvinprovince during the years 2014-2019. The tax file related to guilds is in five tax groups includethe guild group of owners of notary public offices, the guild group of real estate agencies, theguild group of catering halls, restaurants and related businesses, the guild group ofcommunication services, and the guild group of exhibitions and auto accessories stores andrelated businesses. For modeling, the classification model including the decision tree algorithmwas used. The results indicate that the coverage criterion is 68%, the Kappa criterion is 0.612,which indicates the good performance of the modeler. Also, using the Cross Validationtechnique, the validity of the prediction model was tested in order to more reliably estimate thepercentage of modeling performance. The accuracy criterion equal to 67.79% shows theappropriate reliability for the prediction model. The results of this research can be utilized informulating operational strategies based on data mining to predict the tax evasion of guilds in theprovinces.