Intelligent Method for Identifying the Fraudulent Online Stores

Abstract

There is a significant consolidation of IT fraudsters in modern conditions, so the definition of fraudulent sites is a relevant applied study. On the basis of the conducted researches the basic parameters which can be parsed from the required site of online store, namely absence or presence of the corresponding information are defined. An intelligent method for detecting fraudulent online stores has been developed, which allows automating the process of detecting fraudulent sites only by entering a link to the site. The method is implemented on the basis of machine learning classification methods: Logistic Regression (LR), Random Forest (RF), KNN, Naive Bayes (NB), Support Vector (SVM) and DecisionTree (DT). Also, each method of classification is modeled by different approaches, namely: Imbalanced; Undersampling; Oversampling; SMOTE; ADASYN. The realization of method is implemented on the basis of sites operating in Ukraine. There are 67 sites in the dataset, 45% of which are fraudulent.according to the results it turned out that the best simulation estimates are based on the DecisionTree by the approach ADASYN and Random Forest by the approach Oversampling. It shows a 100% result of defining a fraudulent site.