Local community detection based on higher-order structure and edge information

Abstract Local community detection is to discover local community where the seed is located. Most algorithms extend local community by edge information, without considering high-order information in network. The high-order information which is also named as network motif is very important for forming a community. There are also methods that focus on higher-order structure but ignore the sparsely connected edges, resulting in that fail to extend some edge points. In addition, when the seed is the edge node, how to choose the first node to integrate into the community will determine whether the community expands in a right direction. Therefore, a local community detection algorithm based on higher-order structure and edge information (HSEI) is proposed. Firstly, different ways selecting the first node joining local community according to the motif degree of seed are used. Secondly, a new motif-based modularity function is proposed to extend local community, so that the extended community will be connected more tightly. A new motif-based community central node is defined to help extend the central part of local community. For the edge of community and the area with sparse connections, edge information is used to mine the membership strength between nodes and communities, so as to obtain more complete local community members. Compared with five state-of-the-art algorithms, the proposed method achieves better results on the generated networks with different parameters and six real networks.