Abstract: Community search aims to find the local community of a given query node from the network. Community search based on spectrum is one of the popular methods. Existing community search methods based on spectrum are mostly oriented to simple networks, but cannot deal with binary networks with two types of entity association, and the community mining methods oriented to bipartite network are mostly to divide the whole network. Therefore, a spectral approximation community search method for bipartite network is proposed, which aims to introduce spectral method into bipartite network to accurately locate communities closely associated with query nodes. Specifically, firstly, the correlation between two entities in bipartite network is considered, and the local modularity oriented to bipartite network is designed based on the local modularity. Secondly, based on spectral graph theory, a spectral method suitable for bipartite network is designed by using the local approximation feature subspace of modularity matrix fused with different entity associations on bipartite network. Finally, the linear programming problem of sparse indicator vector supported by query node set in spectral subspace is designed by using the local modularity of bipartite network combined with spectral properties. Then the target community can be obtained by solving the linear programming problem. Experiments on real data sets verify the effectiveness and efficiency of the proposed method. 

Key words: bipartite network, community search, spectral approximation, local modularity