Identifying Interaction Groups in Social Network Using a Game-Theoretic Approach

There are often enormous interactions among users of online social networking services. Based on their interactions and the corresponding contents, the users can be clustered into different interaction groups with different interests. In this paper, we propose a game-theoretic approach for formulating the dynamics of users´ participation into interaction groups on a social network formed by users´ interrelationships. A partition of the social network into interaction groups corresponds to an equilibrium of the game. In the proposed approach, the original interaction network is reduced to an effective one based on the attributes of interactions, where the proposed game is played. In the game, each node is represented by a selfish agent that performs actions to join or leave interaction groups according to its utility. An agent may belong to more than one interaction group because of its different interests, which is naturally captured by the proposed approach. The utility of an agent consists a gain function and a loss function. Specifically, we adopt a simple extension to the modularity of a social network as an agent´s gain function and formulate the intrinsic cost paid by the agent for maintaining its interaction groups as its loss function. For the purpose of validation, we conduct experiments on a real-world data set collected from Facebook. The experimental results well demonstrate that the proposed approach is effective in identifying interaction groups in social networks.