From Ants to People: the Vaporization of Social Relationships in Dynamic Community Detection

Community detection in social networks is crucial for understanding social dynamics and interactions. In this paper, we propose LFM2ACO, an algorithm designed to detect dynamic communities by combining the principles of the static Large Families Model (LFM) with those of Ant Colony Optimization (ACO). Inspired by the biological phenomenon of ant colonies, where pheromones guide and reinforce paths to food sources, we model social relationships as dynamic trails that require constant renewal. Just as pheromones evaporate over time unless refreshed by ant activity, social connections weaken without continued interaction. The LFM2ACO algorithm captures this essence, simulating the strengthening of relationships through repeated communication (e.g., messages, likes, comments) and their decay in the absence of such interactions. Comprehensive experiments on real-world social networks, including the Facebook Wall/Links dataset and the Enron email dataset, demonstrate the robustness and efficacy of LFM2ACO in accurately detecting dynamic communities. This work not only enhances the understanding of community evolution but also provides a practical, implemented solution, validated through experimentation, offering valuable insights for future research and development in community detection algorithms.