Label propagation-based approach for detecting review spammer groups on e-commerce websites

摘要: Abstract Online product reviews are very important information resources on e-commerce websites and significantly influence consumers’ purchase decisions. Driven by interests, however, some merchants might hire a group of reviewers working together to promote or demote a set target products writing fake reviews. Such collusive fraudulent reviewer is generally termed review spammer more harmful than individual spammers. To address this issue, in paper we propose label propagation-based approach detect groups websites. First, based the evaluation data reviewers, extract associations between with respect time ratings construct relationship graph reviewers. Second, an improved propagation algorithm intensity automatic filtering mechanism find candidate constructed graph. Finally, ranking that combines entropy method analytic hierarchy process rank thus identify top-k groups. The experimental results real-world Amazon Yelp datasets show proposed performs better baseline methods.