Influence propagation: Interest groups and node ranking models

Abstract

Influence propagation is studied in various contexts with significant practical potential applications such as viral marketing, monitoring people opinions, social psychology analysis and communities discovery. All the previously mentioned applications are concerned about the role played by the user in social network and his/her effect on other users. The current literature lacks approaches that identify influential users in social networks and analyze users ranking with respect to the users interactivity to the disseminated content. The main contribution of this work is to achieve users ranking based on influence propagation in social networks. In order to achieve this goal, two models are proposed. The first model captures interest groups regarding specific disseminated content. The second model is a novel influence propagation model that ranks users in each interest group based on their role in spreading content. Moreover, this model introduces the new concept of ”ultimate observers” to adjust the rank of influential users in each group. Finally, we perform extensive experiments on real datasets to demonstrate the relevance of the proposed models. Both models are evaluated in experimental setup using the following benchmark datasets: Highschool, Email-Eu-core, US Airports, Advogato Trust and Twitter Lists networks. The proposed models are assessed in respect of the accurate separation of interest groups, distinction, uniqueness and effectiveness of nodes ranking. Experiments show that the proposed models have promising results in detecting the interest groups and ranking users in terms of their influence propagation.