Research On Community Detection Algorithm Based On Structural Features Among Neighbors

Many complex systems in reality can be abstracted as complex networks.Communities are an important structural feature in complex networks.Nodes within a community are tightly connected to each other,while different communities are sparsely connected to each other.Studying the community structural features of complex networks is important for exploring and revealing the functional properties of complex systems.The research on community detection algorithms has made great progress,but at the same time,there are many shortcomings.Among them,in the seed-based local community detection method,the seed node selection method,the community expansion method and the community boundary node processing method need further research.In this paper,we use inter-neighborhood structure features to detect communities,and the main research includes the following two aspects.In this paper,we propose a community detection algorithm LEHT based on higher-order triangles from a local perspective to address the problem that closely connected and similar neighboring nodes in a network cannot be reasonably classified into corresponding communities.If the node and its neighbors form multiple higher-order triangles,the stability of the higher-order triangles constructed by the node is calculated using the triangle similarity defined in this paper,and the triangle with the greatest stability is selected as the higher-order triangle of the node.This process is iterated until all nodes satisfying the higher-order triangle have a unique higher-order triangle,and the nodes constructed into higher-order triangles form the initial community.Then,for the node that has a higher-order triangle,the higher-order triangle in which the node is located is merged according to the condition that the node satisfies the condition that there are many common nodes and tight connections between the higher-order triangle to which the node belongs and the higher-order triangle of the neighboring node.This process is iterated until the higher-order triangles satisfying the condition are merged,completing the expansion of the initial community.Finally,for the node that does not have high-order triangles,the proximity centrality of the node’s neighboring nodes in the community is calculated,and the node is added to the community with the largest proximity centrality of neighboring nodes to form the final community.The experimental comparison analysis of the LEHT algorithm with five classical algorithms on the real network and the LFR benchmark network shows that the effectiveness of the algorithm LEHT performs better.To address the problems of local community detection algorithm sensitive to seed node selection and unstable to community expansion,this paper proposes a community detection algorithm HTRS based on higher-order structure and relationship strength.The algorithm first defines a higher-order structure motif degree,divides nodes into higher-order nodes and edge nodes according to the motif degree of each node,and adaptively based on the local neighborhood density of higher-order nodes determine the core nodes of the community.Through multiple higher-order structure motifs formed by the core nodes and their connected close neighbor nodes,the higher-order structure motif with the largest value is selected as the initial community using the relationship strength rs₁.Then,in the higher-order expansion phase,it continues to be subdivided into two expansion methods depending on the number of higher-order nodes connected to the nodes within the community.When two nodes within the community form a higher-order structure motif with a higher-order node outside the community,this higher-order node is extended into the community by the gain function.For a node within the community with two higher-order nodes outside the community constituting a higher-order structure motif,this higher-order node is extended into the community by the relation strength rs₂.Iterate this process until all the higher-order nodes are added to the community.Finally,in the low-order extension phase,the algorithm defines a relation strength rs₃ that considers both the number of nodes in the community to which the edge node is connected and the influence of the node to determine the community to which the node belongs.The edge node is added to the association where the neighboring node with a larger value of relational strength rs3 is located to form the final association.Experimental results on real networks and LFR benchmark networks show that the HTRS algorithm is able to detect associations with high accuracy and stability.In summary,both the LEHT and HTRS algorithms proposed in this paper use the higher-order structural features among neighboring nodes to detect associations by local expansion.the LEHT algorithm solves the problem that closely neighboring nodes cannot reasonably divide associations,and the HTRS algorithm solves the problem of unstable association expansion due to seed node selection.The algorithms in this paper detect associations with high quality and robustness.They provide a feasible solution for detecting association structure.