Direction recovery in undirected social networks based on community structure and popularity

Abstract

Directionality is a significant property of social networks, which enables us to improve our analytical tasks and have a deeper understanding about social networks. Unfortunately, the potential directionality is hidden in undirected social networks. The previous studies on recovering directionality in undirected social networks mostly focus on the microscopic patterns discovered in the existing directed social networks. In this paper, we attempt to recover the directionality based on the macroscopic community structure. To this end, a variant of the existing modularity model, called behavioural modularity, is designed for discovering community membership of nodes. Assuming that members in the same community have higher behavioural similarity, we introduce the concept of the intra-community popularity, and then estimate directionality of undirected ties based on the community structure and the intra-community popularity. Accordingly, we propose a novel Community and Popularity based Direction Recovering (CPDR) approach to recover the directionality of undirected social networks. Experimental results conducted on three real-world social networks have confirmed the effectiveness of the proposed approach on direction recovery.