Research And Implementation Of User Influence Rank Based On Effective Fans

With the rise and development of the Internet era, related social networks sites are becoming more and more popular, such as Sina Weibo, Twitter, Facebook, etc. Each year the growth numbers of users in each social network site have shown rapid trend. Because the actual distances make the contact between people inconveniently, online social networks make users communicate with each other more easily in virtual network. At the same time they can have a significant impact in the way of users to get a different message and share the message. And in today’s tide of the Internet, more and more enterprises and companies want actively to sell and publicize their products widely through social networks, They want to find some small-scale users firstly to accept new products of their company, then attract more users to accept and spread new products by their influence. So how to find the first seed users to use and recommend products to maximize the final product transmission range has become a hot research question. However, many existing user influence rank methods can’t rank user’s influence based on topic, there are some problems of inaccurate result, high time complexity and difficulty for large scale data in these existing methods.Based on the above background and issues, this paper proposes a new parallel user influence rank method UIREF(User Influence Rank based on Effective Fans) in social network based on the definition of effective fans and the concept of themes, it can find different influence users in different theme and make influence spread range larger. Firstly, we define the concept of the number of effective fans in specific theme combining with the theme to measure the number of all users in a particular topic to probably see the published weibo including users who read the original weibo and retweet, so that we can get user’s influence in particular topic and improve the accuracy of the results. Then we achieve a distributed user influence rank algorithm using the Map Reduce model to solve the problem of difficulty for large scale data and improve the speed of the user influence rank, it can be effectively applied in large and complex social networks. Finally, using IC model as a message propagation model, we compare the spread range of users of the method UIREF and other four methods on a number of different sizes of social network data sets, make high influence users ranked by user influence rank methods as initial seed users and verify the effectiveness of the method UIREF. And compare and analyze the run time of serialization UIREF method and parallelization UIREF method to prove parallelization UIREF method can complete user influence rank within a reasonable time on large and complex social networks.