A community discovery algorithm based on local extension of high-order triangle

Abstract

Community structure is an important feature of complex networks. Compared with global community discovery methods, local community discovery methods can discover communities efficiently without complete information about the network structure. Most locally extended algorithms rely on core nodes to discover communities, where the core nodes are based on local density without considering the topological distribution among nodes, and it is difficult to determine the communities to which nodes that are far away and nodes that are in between communities have low similarity to different cores. Therefore, a community discovery algorithm based on local extension of higher-order triangles is proposed at LEHT. The algorithm firstly identifies all three closely linked nodes based on the local topological distribution of nodes in the network using the higher-order triangle structure, and then calculates the most core higher-order triangle of each node using the link strength to form the initial community. Secondly only the core higher-order triangles need to be merged in the topological dimension, and the problem of low similarity of nodes further away from the core is eliminated by extending the direct neighbourhood information between the core higher-order triangles. Finally independent nodes that are not within the core triangles are joined to the community where it is close to a neighbouring node with large centrality. The LEHT algorithm is analysed in comparison with five representative classical algorithms on real and synthetic networks, and the experimental results show that the effectiveness of the LEHT algorithm performs better.