With the advent of big data era,Network Science being focused on complex networks has come into being.An important research area in Network Science is the discovery of community structures.The community structure in a complex network is a cluster with close internal connections and sparse external associations.As a bridge to understand the relationship of individual and whole in network,community structure,which is a key hub to implement predictive control,plays an essential role in recognizing the microscopic and macroscopic structure,as well as the temporal and spatial evolution in complex network.Existing methods for discovering complex network community are mostly concentrated on non-overlapping network structures,and rarely have overlapping community detection methods matching the actual network well appeared.Chaotic complex structure makes mining hard,and overlapping nodes will be unchanged after detecting.Consequently,Overlapping community detection algorithm based on Bi-Partition network(OBP)is proposed.OBP transforms the original network into a clearly Bi-Partition network innovatively,uses label propagation to implement link clustering novelly,determines overlapping nodes again concisely,and completes partition density metric tersely.Furthermore,Overlapping community detection algorithm by applying random walk based on Markov Chain cLustering(OMCL)is also proposed to reconstruct the network according to the matrix reached by the expansion for the purpose of information flow and the inflation in the way of strengthening the community and weakening the frontier.In this paper,OBP being robust is compared with three state-of-the-art algorithms on four artificial datasets and eight real networks.The experiment proofs that our method has significant performance in EQ and EPD respectively,and the result is stable.Simultaneously,the four artificial datasets and seven real networks are used to compare the performance of OMCL with other five excellent methods.In all networks,the comprehensive performance of OMCL in PD and EQ is superior,and the algorithm converges quickly. |