Social network analysis: a novel paradigm for improving community detection

Abstract

Social network analysis has become increasingly important across a wide range of fields, offering valuable insights into complex systems of interconnected entities. One of the fundamental challenges in this field is the community detection problem, which involves identifying groups within networks. Multiple algorithms have been proposed, exploring new approaches to finding solutions for cohesive partitions of the graph. One of the most considered philosophies when defining this type of technique is the use of the graph’s adjacency matrix as input and the consideration of modularity as the function to be optimized. We propose an enhancement to this approach to community detection by incorporating high-order relationships between nodes, allowing for a more comprehensive capture of network structure. By modifying the algorithm’s input, our method improves community detection accuracy. Moreover, our proposed approach is universal, applicable to any algorithm that utilizes a matrix as input. Its value is further validated through a comprehensive set of results, comparing the original problem with the enhanced method we present. We also present a tourism case study.