Detecting spammer groups using centrality measure

Abstract

The practice of hiring professional spammers to create fake reviews on e-commerce platforms in order to attract customers can easily be found on the internet. As a result, the number of suspicious spammers has rapidly increased, hindering the performance of fraud/spam specialists' investigations. This study proposes a solution to mitigate the overwhelming number of investigative cases and improve investigation performance. The study utilizes XGBoost for classification to detect individual spammers; and centrality measure and the Structural Clustering Algorithm for Networks (SCAN) to identify spammer groups. Experimental evaluations show that this approach effectively reduces the number of spammers and increases accuracy and precision compared to previous studies, particularly concerning the hundred highest suspicious groups.