Community Detection Algorithms Based On Modularity Optimization

Complex networks are widely used in complex systems research in the fields of social and biological fields.Analyzing the Existential Community Structure in Complex Networks is important.A lot of existing methods used marginal probability based on modularity to assign each node to its most likely community.However,selecting the greatest marginal probability of a node in which communities,leads to poorly correlated with each other.Moreover,there ate a lot of real-networks change dynamically,and most of its are directed networks.So We propose different algorithms.Firstly,for the undirected static complex network,we propose a method that tread modularity as energy function and used an efficient Belief Propagation method to obtain the consensus of many partition with high modularity.In addition,One of the technologies we used could reduce the time complexity to the linear from quadratic level.Testing on networks showed the applicability of our method superior performance over competing methods in community accuracy.For the directed dynamic network structure,an incremental DNGI algorithm for directed network community discovery algorithm is proposed.The experimental results on a real directed network show that our algorithm performs very well both in accuracy and efficiency.In summary,we design a community discovery algorithm that uses belief propagation for a static network.The algorithm can find a globally optimized community.On the other hand,for the directed dynamic network structure,an incremental community detection algorithm for directed networks is proposed.The change of the community results can divide the community structure in real time.