Method of Detecting a Fictitious Company on the Machine Learning Base

The role of fictitious firms, which are also conventionally called “anonymous commercial structures”, is very high in committing economic crimes. The problem of their detection is complicated by the fact that they break the chains of economic and financial relations between real enterprises, banks, insurance companies, exporters and importers, trade enterprises. To solve this problem, the authors have developed a method based on machine learning, which detects fraudulent activities of a fictitious enterprise (sham business). Based on the analysis of the existing relevant publications the parameters of detecting the fictitious enterprises are formed. As a result of modeling the economic activity of 100 Ukrainian enterprises, 20 fictitious enterprises were identified by the methods of Support Vector Classifier, Stochastic Gradient Decent Classifier, Random Forest Classifier, Decision Tree Classifier, Gaussian Naive Bayes, K-Neighbors Classifier, Ada Boost Classifier, Logistic Regression. The results of experimental studies have shown that all selected methods of classification have an acceptable result. However, the best are Support Vector Classifier, Gaussian Naive Bayes, Logistic Regression with a forecast score of 0.98 and a standard deviation of 0.02.