Overlapping Community Detection Based on Weak Equiconcept

Community discovery refers to the process of searching for clusters in a network that are sparsely connected to other nodes and formed dense connections internally. In many real networks, some communities often overlap with each other, meaning that a node may belong to multiple communities simultaneously. Revealing these (overlapping) community structures is an important issue in complex network analysis, as it helps to better analyze the characteristics and organizational structure of the network. Community expansion methods are very important in the study of community detection problems. However, one of the key issues in developing effective community expansion methods is that the position of seed nodes greatly affects the performance of these algorithms, resulting in low robustness of these algorithms. Meanwhile, it is also difficult for them to provide high-quality results for community detection task. To solve the above problem, this paper proposes a seed selection method based on the weak equiconcepts in a network formal context, which integrates the attribute information of nodes during random walk to detect overlapping communities. Specifically, the weak equiconcepts are constructed by establishing a network formal context to obtain seed sets, and an improved PageRank clustering algorithm is used to expand these seed sets to better reveal the overlapping community structure in the network. Experiments show that seed selection is helpful to improving the performance of overlapping community detection algorithms.