Common-neighbor based overlapping community detection in complex networks

This study tackles the problem of detecting overlapping communities in complex networks, a task that continues to be challenging due to its NP-hard nature. To tackle this issue, existing overlapping community detection methods face substantial challenges, including accommodating numerous community memberships, effectively differentiating between overlapping and non-overlapping nodes, and most of the time not paying attention to the subtle communities. Thus, developing effective community detection methods that accommodate these diverse scenarios is essential for understanding and analyzing complex network phenomena. Therefore, we propose an approach to overlapping community detection in complex networks and suggest a method that uses common neighbor (of two nodes is a node adjacent to both) or neighbor similarity as a crucial parameter that has an effect throughout the process. To benchmark our algorithm against existing ones, we choose some top-performing algorithms from the Community Detection library, filtering out those that rely on prior knowledge, exhibit instability, or consume significant computational time. We assess the effectiveness of both the proposed algorithm and the selected algorithms by employing a range of established metrics, including modularity, F-score, and Normalized Mutual Information. Additionally, we incorporate overlapping communities into the coverage metric that was created for disjoint communities and use this metric to perform comparative analyses. To evaluate the performance of each algorithm, we conduct further examinations on small graphs representing realistic communities. The proposed approach successfully detects overlapping communities, as demonstrated by the experimental results.