Risk assessment of VAT entities using selected data mining models

Abstract

The goal of the paper was to evaluate the classification ability of selected types of data mining methods, focusing on neural networks, decision trees and random forests, within the risk assessment of VAT entities. The data set used for the testing contained information on the risk of taxpayers who were obliged to file VAT returns in the calendar year 2012. The highest classification ability among the constructed models was achieved by the multilayer perceptron model. The lowest classification ability was demonstrated by the decision tree method, using the default growth exhaustive CHAID algorithm.