FuSeO: Fuzzy semantic overlapping community detection

Abstract

Finding community structures in online social networks is an important methodology for understanding the internal organization of users and actions. Most previous studies have focused on structural properties to detect communities. They do not analyze the information gathered from the posting activities of members of social networks, nor do they consider overlapping communities. To tackle these two drawbacks, a new overlapping community detection method involving social activities and semantic analysis is proposed. This work applies a fuzzy membership to detect overlapping communities with different extent and run semantic analysis to include information contained in posts. The available resource description format contributes to research in social networks. Based on this new understanding of social networks, this approach can be adopted for large online social networks and for social portals, such as forums, that are not based on network topology. The efficiency and feasibility of this method is verified by the available experimental analysis. The results obtained by the tests on real networks indicate that the proposed approach can be effective in discovering labelled and overlapping communities with a high amount of modularity. This approach is fast enough to process very large and dense social networks. © 2017-IOS Press and the authors. All rights reserved.