FAKE REVIEWS DETECTION BASED ON BOTH THE REVIEW AND THE REVIEWER FEATURES UNDER BELIEF FUNCTION THEORY

The online reviews play an increasingly spreading role in consumer purchasing decisions and they are also considered as one of the most powerful source of information for companies. Due to this attraction, manufacturers and retailers rely on spammers to promote their own products and demote the competitors’ one by posting fake reviews. Therefore, it is essential to detect deceptive reviews in order to ensure customers confidence and to maintain companies' fair competition. To tackle this problem, we propose a new approach able to spot spam reviews relying both on the rating reviews and the different spammers' indicators under the belief function framework. This method treats uncertainty in the given reviews also in the reviewers' information to take into account each reviewer spamicity when making decision. Experiments are conducted on two real-world review data-sets from Yelp.com with filtered (spam) and recommended (non-spam) reviews to demonstrate our method the effectiveness.