Identifying influential users by their postings in social networks

Abstract

With the rapid development and increased popularity of social networks, much research effort has been conducted to analyze information of social networks, such as finding the influential users. Our research is focusing on identifying the influential social network users; as it can help to increase the marketing efficiency, and can also be utilized to gather opinions and information on particular topics as well as to predict the trends. Different from previous work, our aim is to identify the most influential users based on the interactions in their posts on a given topic. We first propose a graph model of online posts, which represents the relationships between online posts of one topic. Three measurement methods have been developed to assess the influences of posts, so as to find the influential posts on the topic. In our work, there are two types of influences based on the different roles: starter and connecter. A starter is followed by many others, similar to a hub in a network, so it should have certain influence. A connecter is also regarded to be influential when it links starters together. After we can measure the influence of online posts and find the influential posts, their authors can be considered as potential influential users. With the consideration of some users would have several influential posts, we develop a user graph model to refine the influence measures to find influential users. Based on the authors of influential posts found, we convert a post graph to the corresponding user graph, and then measure the influence of users which are starter and connecter respectively. Finally, the most influential users can be determined in the user graph. Also, our proposed model can be extended and used to find the sentimental influence of posts and users. We have conducted two case studies in order to verify our proposed graph models and influence measurement methods. In the first study we applied the graph model of online posts and visualized the result of starter and connecter identifications. The experiment is performed on Twitter, and it shows that the influential starters and connecters in the post graph can be identified after integrating the results from three measurement methods. In order to validate our model, we compared the results of our methods with three centrality metrics and the PageRank algorithm. The experiment result shows that our proposed methods outperformed the others in the ability of identifying both starters and connecters. Next, the influential users identified by the post graph model and the user graph model are compared in the second case study. The results show that users with more influential posts may not be truly influential, when the followers are always the same group of people. In a user graph obtained from a given post graph, the connecters already identified in the original post graph can be refined and some new ones found are considered to be potential connecters.