Overlapping Community Detection in Complex Networks

Complex networks, from social media graphs to biological interactomes, often exhibit community structure in which nodes cluster into groups with denser internal connections. Unlike disjoint partitions, overlapping community detection acknowledges that nodes may participate simultaneously in multiple functional modules. This reflects real-world scenarios such as individuals belonging to several social circles, proteins involved in multiple pathways, or articles spanning diverse topics. Methods for uncovering these overlaps range from extensions of classical node-based clustering to link-centric and hybrid schemes. Key challenges include defining clear membership criteria, balancing computational efficiency against detection accuracy, and revealing hierarchical organisation. Advances have integrated ideas from graph theory, statistical inference and optimisation to yield scalable, robust algorithms capable of handling weighted, directed and dynamic networks. As the field matures, attention is shifting towards methods that combine local and global information, offer principled model selection and provide insights into the functional significance of overlaps.

Research from Nature Portfolio

Recent seminal work has introduced network decomposition approaches that iteratively remove and reassemble link communities derived from node clustering. This strategy reduces noise and computational cost, while avoiding ambiguous link-similarity measures. It has been demonstrated on both synthetic benchmarks and real-world datasets to yield high accuracy in recovering overlapping modules. Another breakthrough employs edge label propagation to uncover both link-based and node-based communities in a unified framework. By allowing edges to carry labels that diffuse through the network, this method captures pervasive overlaps with linear time complexity, making it suitable for very large systems. A complementary development uses random-walk-based seed expansion: initial high-degree nodes define seed communities, probabilities of node affiliation are computed via random walks, and an optimisation step refines boundaries. This approach excels in networks with weak or ambiguous community delineations, outperforming several state-of-the-art algorithms in quality metrics.

Research from all publishers

A 2024 algorithmic advance employs hierarchical agglomerative clustering on maximal cliques, introducing novel dissimilarity measures based on closed-trail distance and overlap size. This deterministic method naturally generates nested overlapping communities and outperforms conventional hierarchical and flat algorithms on benchmark networks. In 2021, a local expansion strategy was refined by selecting seeds through a fusion of node degree and clustering coefficient using an entropy-weight scheme. Adaptive growth functions then expand communities, followed by merging and isolated-node adjustment to produce high-precision partitions. A 2020 comparative analysis surveyed representative overlapping detection methods across synthetic and ground-truth networks, revealing that many algorithms miss critical structural properties such as overlap region size and multi-membership patterns. This work highlighted the need for evaluation metrics beyond standard quality scores, guiding the design of next-generation algorithms.

Overlapping Community Detection in Complex Networks publication trend

The graph below shows the total number of articles in overlapping community detection in complex networks across all publications each year (not limited to Nature Index journals).

Technical terms

Overlapping community: A set of nodes in which each node may belong to more than one group, reflecting shared or multifaceted relationships.

Clique: A subset of nodes all pairwise connected, often used as a building block for detecting dense local structures.

Link community: A community defined by grouping edges rather than nodes, capturing overlap through shared links.

Network decomposition: A process of iteratively splitting and reassembling a network to reduce noise and simplify community identification.

Label propagation: An iterative algorithm where nodes or edges adopt community labels from neighbours to reach consensus on membership.

Seed node: An initial node chosen to initiate a community expansion process, often selected for high centrality or connectivity.

References

1. A hierarchical overlapping community detection method based on closed trail distance and maximal cliques. Information Sciences (2024).
2. Overlapping Community Detection based on Network Decomposition. Scientific Reports (2016).
3. Discovering communities in complex networks by edge label propagation. Scientific Reports (2016).
4. A seed-expanding method based on random walks for community detection in networks with ambiguous community structures. Scientific Reports (2017).
5. A Local Extended Algorithm Combined with Degree and Clustering Coefficient to Optimize Overlapping Community Detection. Complexity (2021).
6. A comparative study of overlapping community detection methods from the perspective of the structural properties. Applied Network Science (2020).