Community detection is a key task in the analysis of complex networks, particularly in social network analysis, where uncovering cohesive and well-separated groups is essential for understanding structural organization and interaction patterns. Many existing centroid-based community detection methods rely primarily on node degree for centroid selection, which often leads to centroid crowding and insufficient spatial separation among communities. To address these limitations, this paper proposes Degree–Distance Centroid–Community Detection with Influential Nodes (DDC-CDIN), a distance-driven and influence-aware community detection framework. In the proposed approach, nodes are first ranked according to an Enhanced Degree Centrality measure that incorporates degree information, neighbourhood structure, and local clustering characteristics to identify structurally influential nodes. Centroids are then selected iteratively from the top-ranked influential nodes by maximizing shortest-path distances, ensuring that the chosen centroids are both representative and well dispersed within the network. Once the centroids are determined, the remaining nodes are assigned to communities based on the minimum geodesic distance, yielding compact, clearly separated clusters. Extensive experiments across multiple real-world networks show that DDC-CDIN achieves competitive performance compared to traditional centroid-based and modularity-driven methods in terms of modularity, community cohesion, and boundary clarity. The results indicate that jointly incorporating influence-aware node ranking with distance-based centroid dispersion effectively mitigates centroid crowding and enhances overall community detection quality. These findings demonstrate the effectiveness and robustness of DDC-CDIN for detecting well-structured and topologically coherent communities in complex networks.
Amedapu et al. (Mon,) studied this question.