Review Spam Detection Using Semi-supervised Technique

Today because of the popularity of e-commerce sites, spammers have made their target to these sites for review spam apart from other spams like email spam or web spam. These fake reviews written by fraudsters prevent customers and organizations reaching actual conclusions about the products. Hence, these must be detected and eliminated so as to prevent deceptive potential customers. In this paper, we have used semi-supervised learning technique to detect review spam. The proposed work is based on PU-learning algorithm which learns from a very few positive example and unlabeled data set. Maximum accuracy we have achieved is of 78.12% with F-score 76.67 using only 80 positive example as a training set.