Fraudulent e-Commerce website detection model using HTML, text and image features

Abstract

Many of Internet users have been the victims of fraudulent e-commerce websites and the number grows. This paper presents an investigation on three types of features namely HTML tags, textual content and image of the website that could possibly contain some patterns that indicate it is fraudulent. Four machine learning algorithms were used to measure the accuracy of the fraudulent e-commerce websites detection. These techniques are Linear Regression, Decision Tree, Random Forest and XGBoost. 497 e-commerce websites were used as training and testing dataset. Testing was done in two phases. In phase one, each features was tested to see its discriminative capability. Meanwhile in phase two, these features were combined. The result shows that textual content has consistently outperformed the other two features especially when XGBoost was used as a classifier. With combined features, overall accuracy has improved and best result of accuracy recorded was 98.7% achieved when Linear Regression was used as a classifier.