Graphical Framework For Review Spammer Group Detection Using Metadata

Nowadays, many people depending on online reviews for the purchasing decision of a product/ service. One of the characteristics of an online review system is that anyone can post a review that allows spammers to compose fake reviews. Recently, these spammers work as groups to intensify their activities and for maximum profit gains. Few works are concentrated on the detection of group spammers compared to individual review/reviewer spamming. This work proposes a framework to detect spammer groups using graph- based algorithms with five group spamming features and also proposes a new group spamming feature, Group Rating Similarity (GRS) based on the review rating score. The results show that the proposed framework performs well with five features when comparing with the existing work having seven features. Also, the proposed feature (GRS) shows better performance in discriminating spam and non-spam when experimented on realworld review datasets from the Yelp website.