Identifying Spammers by Completing the Ratings of Low-Degree Users

Along with the rapid development of e-commerce, a large number of spammers disrupt the fair order of the e-commerce platform. The false ratings rated by these spammers do not match the quality of items, confusing the boundaries of good and bad items and seriously endangering the real interests of merchants and normal users. To eliminate the malicious influence caused by these spammers, many effective spamming detection algorithms are proposed in e-commerce platforms. However, these algorithms are ineffective in judging how trustworthy a user with insufficient rating data. In order to address this issue, we take inspiration from traditional recommender systems by completing the missing ratings of low-degree users to improve the efficiency of spamming detection algorithms when approaching those users. User similarity is used in this paper to predict the missing ratings of users. A novel reputation ranking method is proposed. We then test our improvements compared with DR, IGR, and IOR. Experimental results on three typical data sets suggest that our method combined with IOR has improved by at least (formula presented) in dealing with malicious spammers, respectively. As for results on detecting random spammers, our method improves by a least (formula presented), respectively.