A Novel Fusion Approach Based on RF and XGB for Detecting Fake Reviews on E-Commerce Platforms

The proliferation of fraudulent reviews on e-commerce platforms is exacerbated when unscrupulous firms and service providers enhance their ability to manipulate consumer behaviour through the use of counterfeit evaluations. Bot farms, Sybil accounts, and the use of legitimate accounts to promote products and disparage competitors are prevalent elements of these fraudulent activities. The research community has been compelled to reevaluate existing detection approaches due to the intricate nature of review spamming, which significantly influences purchasing decisions, brand reputation, and vendor responsiveness. This paper introduces the RFXGB hybrid machine learning algorithm for the precise detection of fraudulent reviews. Data preparation entails the application of several procedures to raw review data to enhance its usability. Subsequently, to accurately represent textual material, feature extraction employing TF-IDF is implemented. The model's effective classification is attained by integrating an XGB classifier with an RF for feature selection. The RFXGB model demonstrates significant potential in identifying fake information, achieving a detection accuracy of 96.48%, as evidenced by experimental assessments employing several performance metrics. The findings indicate that the model enhances the reliability of online review systems and that advanced, adaptive models are essential to address the issue of fraudulent reviews on e-commerce platforms.