Mining of Attribute Network Community based on Structure and Node Similarity

Abstract

Mining cohesive subgraphs from a network is an important direction in network analysis. Most of the existing methods are based on the topology of common networks, which ignores the rich information of the nodes in the real network. The k-truss model which is proposed by user engagement and tie strength, captures the degree of strong connection among users who participate in the network with other users actively. However, this model does not consider the attributes of users. In order to find the cohesive subgraphs on social networks efficiently and accurately, this paper proposes a new model (k,r)-truss on the attribute network community based on k-truss, and finds the cohesive subgraphs on the social network from the perspective of strong connection and similarity between users. The problem of enumerating all maximal (k,r)-truss is NP-hard, so in order to speed up the calculation, this paper proposes new pruning algorithms AdvEnumH and AdvEnumHC, which reduces the search space of the mining process significantly. Finally, the experiments are carried out on the real data set to evaluate the performance of the proposed algorithm. The results of experiments demonstrate that our algorithm has significantly improved in efficiency and timeliness compared with the current best method.