Online Detection Of Fake Product Reviews

With the rapid development of Web 2.0 technology in the Internet era,users have regarded online product reviews as an important reference resource for selection of products and services.Some businesses or individuals are motivated by interests to intentionally make review spam,promote their products by recruiting review spammers,or denigrate the products of competitors.The behavior harm of review spammers group is greater than that of single review spammers,and the characteristics of review spammers group are more obvious than that of single review spammers.Therefore,the main form of detection of review spam cheating is to detect the review spammers group.This paper presents an online detection method for fake product reviews.In the first stage,the time window mechanism is used to extract the review data in the time window to generate two parts of the relationship between critic and product.Then,the relationship between critic and critic in the two parts of the graph is projected to generate the review figure.Finally,the weight of the edge between reviewers in the reviewer diagram is calculated,and the edge with weight greater than or equal to threshold value is retained.Due to the action of the time window mechanism,the weighted review graph generated in the time window becomes a dynamic weighted reviewer graph.In the second stage,the weighted reviewer graph generated in the time window was clustered by SCAN algorithm to extract the group of candidate false reviews.In the third stage,Image clustering was perforled by SCAN algorithm to extract candidate review spammers group.In the fourth stage,The SVM was used to model the fraud characteristics of candidate review spammers group.The experimental results show that by applying the online detection method of fake product reviews proposed in this paper,high precision rate and recall rate can be obtained,which is of great significance to the detection of false product reviews in the review data set.