Research Issues in MSNs

There have been many efforts to improve the performance of MSNs. We discuss several important research directions in the following subsections, i.e., community detection, content distribution, mobility, privacy and security.

Community Detection

Community detection is to discover communities within which users have the common interests or close social relationship. Some existing methods [32, 33] can be used to divide MSNs into some communities. However, as most of these methods are not designed based on users’ social features, they are still inefficient to detect the unknown communities in MSNs. Furthermore, due to the mobility of users and the large scale of MSNs, the community detection in MSNs becomes more difficult. Therefore, there have been many studies for the community detection in MSNs.

Yang et al. [34] propose a semi-supervised community detection framework, in which the prior information is used to penalize the latent space dissimilarity of different nodes to detect communities in a social network. The proposed framework can improve the accuracy of community detection for networks with unclear structures. Lu et al. [35] focus on the community detection in weighted networks and propose a community detection method based on the strength of the relationship between users in the network. The authors define the intra-centrality and inter-centrality to characterize relationship between users in the network. Jia et al. [36] define the edge centrality to discriminate the edges between networks nodes. Based on an edge centrality, a novel algorithm called edge antitriangle centrality is proposed for community detection.

Whang et al. [37] propose an overlapping community detection method based on a seed expansion approach. The proposed method can not only produce cohesive clusters but also identify ground-truth communities by greedily expanding seeds with a community metric. Chen et al. [38] discuss two fine-tuned community detection algorithms to measure the community quality where the given network community structure can be split and merged. Su et al. [39] propose a fuzzy modularity maximization method for overlapping community detection, which is based on a modularity maximization algorithm. The globally optimal solution can be found by a tree-based structure.

< Prev   CONTENTS   Source   Next >