Research On Community Detection Methods Of Bipartite Networks

Bipartite network is a kind of important complex networks.Its particularity lies in there are two kinds of nodes,not just as one mode network consists of one class of nodes.It contains two kinds of nodes,and the edges between two kinds of nodes,and no edge between the same kinds of nodes.Bipartite network often occurs in real life,it reflects the real and objective characteristics of our life as a type of network.Therefore,the study of bipartite network becomes very meaningful and particularly valuable.Community detection is a hot spot of complex network analysis,is also an important analysis methods of network information mining.Bipartite network community detection had preliminary progress in the past decade,but its theoretical system is not perfect.This article is mainly studied the following two aspects of the bipartite network community detection in-depth research and experiment,innovative results are as follows:(1)Through classification and sorting the algorithm of bipartite network community discovery,overall,can be divided into the overall community detection and classification detection,and make in-depth analysis and discussion of these algorithms.A common problem in comm-unity detection is that it is difficult to determine the number of community,and make a further study.It is proposed a new algorithm CAA(Clustering Allocation Algorithm).Firstly,start from one class nodes of the network,find out the two nodes with the least similarities.Secondly,taking out the nodes with similarity is greater than a threshold to the two selected nodes.Repeat the above steps until the class nodes all are removed.At this point,we find the number of community;Will be Then according to the definition of the intimacy of index in this paper allot another kind of nodes to the existing community.So we get the division of the community structure of the bipartite network.This algorithm can not only independently determine the number of the community,but also can efficiently find the community structure.Experiments show that the algorithm in artificial data and real data show the good effect.(2)The algorithm of bipartite network community discovery has achieved initial progress,but many algorithms are complicated to operate,and contains parameters.Based on the above problems,this paper proposes a simple and strong maneuverability,based on the nonnegative matrix factorization of the bipartite network community detection algorithm.Firstly,the algorithm establish the objective function of optimization model based on the nonnegative matrix factorization,community index matrix of two types of nodes are obtained after the iterative formula update are obtained by the block coordinate descent method,and find the community detection of the bipartite network.Experiments show that the algorithm in artificial data and real data show the high efficiency.