Detecting Community Evolution by Modeling Individual Traits

Abstract

The detection of community structures in social networks has become an increasingly popular topic of research. A particular challenge lies in tracking the evolution of these structures, along with the accompanying community semantics, over time. Addressing this challenge requires a comprehensive understanding of the underlying factors that drive community evolution. This study contributes to that understanding by emphasizing the critical role of individual traits in influencing community dynamics. Building on this insight, we propose an innovative model for dynamic community detection that incorporates inherent evolutionary features. The model segments the social network into temporal snapshots and identifies community structures by leveraging both individual traits and information derived from preceding snapshots. Simultaneously, it tracks evolving discussion topics within communities, enabling the modeling of community semantics. Additionally, the model provides a robust framework for user trait profiling. Its effectiveness is validated through extensive experiments on three real-world datasets. Compared with seven state-of-the-art baselines, our model consistently demonstrates superior performance. Case studies further confirm its capability to uncover community semantics and accurately characterize user traits.