Community Search and Graph Structure Analysis

Community search and graph structure analysis encompass a suite of computational methods designed to identify and characterise cohesive subgraphs within large networks. The chief aim of community search is to retrieve a subgraph that best captures a set of query nodes or fulfils a user’s criteria of cohesion, density or influence. Graph structure analysis more broadly examines the organisation of networks through hierarchical decompositions—such as k-core and k-truss—that reveal layers of connectivity, resilience and functional modules. These approaches underpin applications ranging from the detection of protein complexes and social circles to the optimisation of information flow and infrastructure robustness. Advances in model formulation, algorithmic efficiency and attribute integration have expanded the scope to multi-valued networks, dynamic settings and personalised queries, thereby enhancing both theoretical understanding and practical utility on real-world data.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

One foundational approach introduces the k-plex community model for query-driven search. In this framework, communities are defined as maximal subgraphs in which each vertex may lack connections to at most k other vertices, striking a balance between clique-like strictness and k-core flexibility. Exact and heuristic algorithms based on branch-and-bound techniques allow identification of optimal k-plexes containing a given set of query nodes, addressing the computational hardness via effective pruning and fast candidate generation.

In multi-valued networks—where nodes carry multiple attributes—a skyline community concept has been proposed. Here, maximal k-cores that are non-dominated with respect to all numerical attributes are extracted, capturing communities that cannot be simultaneously outperformed on any attribute. Dimension-reduction methods based on maximum entropy and specialised pruning algorithms enable efficient enumeration of such skyline communities, even when attributes exhibit duplicate values across nodes.

Addressing the need for personalised relevance, maximal personalised influential community search has been formulated. Each node is assigned an influence score, and the task is to find a k-core containing a query vertex that maximises aggregate influence. Top-down and bottom-up algorithms, complemented by index-based strategies, support rapid online queries and retrieval of top-r communities, thus catering to scenarios with multiple, repeated queries in large networks.

Community Search and Graph Structure Analysis publication trend

The graph below shows the total number of articles in community search and graph structure analysis across all publications each year (not limited to Nature Index journals).

Technical terms

Community search: Retrieval of a subgraph that best aligns with specified query nodes or cohesion criteria.

k-core: A maximal subgraph in which each vertex has at least k neighbours within that subgraph.

k-plex: A relaxation of clique structure; a subgraph where each vertex is adjacent to all but at most k other vertices in the subgraph.

k-truss: A cohesive subgraph in which every edge participates in at least k−2 triangles within the subgraph.

Skyline community: A maximal k-core that is not dominated by any other community on all node attributes in a multi-valued network.

Influence value: A numerical score assigned to each node, used to rank or weight nodes in community extraction.

References

1. Query Optimal k-Plex Based Community in Graphs. Data Science and Engineering (2017).
2. Fast skyline community search in multi-valued networks. Big Data Mining and Analytics (2020).
3. Efficient Personalized Influential Community Search in Large Networks. Data Science and Engineering (2021).