A Method for User Credibility Evaluation on Online Business Review Services

The review mechanism provided by e-commerce services plays a critical role in the decision-making process of online consumers. Due to its importance, the review mechanism has also been misused for malicious intents, such as spamming and posting fake reviews. Although many innovative techniques have been developed to filter spam reviews and spammers, these techniques could not help users further pick out high-credible information. In this paper, we proposed a machine-learning-based approach to evaluate the credibility of reviewers. Typically, high-credible reviewers are more likely to post reliable reviews. Therefore, our approach can help users discover high-credible information more efficiently. We propose three groups of features that measure reviewer credibility in terms of user behavior, social relationships, and review quality. Then, we learn a classifier with a resampling mechanism to decide whether a user is high-credible or ordinary. Different from existing studies, which mostly focus on online shopping websites, our targets are users on business review services, such as Yelp and Dianping.com. Experimental results have demonstrated the effectiveness and advance of our approach.