Abstract
Community and hierarchical structures are frequently created, disrupted, and reorganized in dynamic social networks. Understanding these dynamic processes is useful for a wide range of applications, including information diffusion, social innovation, and organizational management. In this research, we conduct a comprehensive evaluation of hierarchical reconstruction and community identification on three different datasets: the Email-EuAll network, Facebook Social Circles, and a synthetic dataset that simulates real-world network behavior. We assess the effectiveness of many well-known community detection techniques, including Louvain, Walktrap, Clique Percolation, and Label Propagation, in order to determine how well they recognize dynamic network patterns. Our findings demonstrate that while Louvain and Walktrap are successful in modularity-based scenarios, effectively detecting well-defined, densely interconnected communities, Label Propagation performs better in sparse networks with fewer, loosely coupled communities. Furthermore, Clique Percolation struggles in extremely dynamic environments yet provides valuable insights into overlapping community structures. Given that different algorithms yield varying degrees of accuracy and efficacy in dynamic situations, the results demonstrate that the best method to employ depends significantly on the structural characteristics of the network. The adaptable nature of hierarchical structures is further supported by empirical studies, which highlight how community patterns shift as a result of real-time network changes. The need for dynamic modeling techniques that can adjust to shifting network dynamics over time is supported by these findings.
Keywords: Community, Community Detection, Dynamic Social Network, Hierarchy, Network Analysis, Social Network Analysis.