GSCPM: CPM-Based Group Spamming Detection in Online Product Reviews

摘要: Online product review is becoming one of important reference indicators for people shopping, but the current site contains a lot fraudulent reviews. Group spamming, which involves group reviewers writing reviews or more target products, becomes main form spamming. However, solutions spammer detection are very limited, and due to lack ground-truth data, this problem has never been completely solved. In paper, we propose novel three-step method detect spammers based on Clique Percolation Method (CPM) in unsupervised way, called GSCPM. First, it utilizes clues from behavioral data (timestamp, rating) relational (network) construct suspicious reviewer graph. Then, breaks whole graph into k-clique clusters CPM, consider such as highly candidate spammers. Finally, ranks groups by group-based individual-based spam indicators. We use three real-world datasets Yelp.com verify performance our proposed method. Experimental results show that outperforms four compared methods terms prediction precision.