An Approach To Reviewer Spammer Group Detection Based On SCAN

In recent years,with the rapid development of Internet technology,a new situation has been bought for the online product and customer.The reviews of product determines whether customers choose to buy.The evaluation of products has a crucial influence both for the customer to make choices and the online store to make profits.More and more fake reviews on the online store and the fake reviewers become a trend,which caused in the discussion and study.In this paper,we discuss the review spammer group detection problem.We present a formula to calculate co-reviewing collusiveness,and generates the weighted review spammers graph.Using SCAN algorithm generate reviewer spammer group based on the weighted review spammers graph.The new group spam indicators are introduced in this paper based on previous research,mining spam indicator from an individual behavior point,and take the average as one spam group indicator to calculate.In the final experimental phase,we take two datasets to make the comparison experiment.According to the experimental results,analyzes the advantages and disadvantages of this algorithm.At the same time,comparing this algorithm with the GSBP algorithm,in the two dataset,the accuracy of this algorithm is better than that of GSBP algorithm.We also make the artificial evaluation,we select 100 groups from the result of our algorithm and make hand marking labels,and comparing with the original label of the dataset and get a higher accuracy.