A Thematic Analysis and Detection of Fraudulent Online Ratings

This study examines the characteristics and detection of fraudulent online reviews using different state-of-the-art machine learning algorithms like support vector classifier (SVC) and logistic regression (LR) and random forest (RF). We analyzed online reviews across ten categories. Our analysis of thematic content of these reviews revealed an overlap in commonly used words between legitimate and fraudulent online reviews. However, less frequent words offered better differentiation. Legitimate reviews used context-specific terms, while fraudulent reviews lacked such specificity. This suggests focusing on the “long tail” of less fraudulent review words for detection. In terms of fraudulent detection, the SVC outperformed the RF and LR by obtaining the highest accuracy (89%) and F1-score (89%). These findings highlight the importance of analyzing uncommon words for fraud detection and effectiveness of machine learning, particularly SVC, for this task. The study results show how crucial linguistic patterns and machine learning can offer consumers and review platforms tools to safeguard user experience and promote online safety and trust.