Mining User Opinion Influences on Twitter Social Network: Find that Friend who Leads your Opinion Using Bayesian Method and a New Emotional PageRank Algorithm

With about 326 million[1] monthly active users and many millions of tweets sent per day, Twitter is undoubtedly one of the social networks most requested[2] by users sharing opinions, and feelings about trends, events… As a result, how users influence their opinions mutually constitute a hot issue for researchers. Indeed, the study and the estimation of the opinion influence observed between Twitter users constitutes a rich opportunity for the adjustment of services/products offered involved in the service discovery process. In this paper we propose an approach to determine the target user's Twitter friends from whom the target user opinion is influenced by. Our model is based on opinion mining of retweets and target user's Favorites markings from which we estimate the opinion influence using the Bayesian method combined with our EPR (Emotional PageRank) algorithm. The results obtained highlight our contribution compared to the standard PR (PageRank) algorithm.