Research On Community Detection Of Complex Network Based On Modularity Strength

In our earth,there are various kinds of relationships,most of which can be extracted as the form of complex networks.Up to now,complex network has obviously become one of the most important parts of data mining and been attracting more and more researchers' attention.After a lot of research and exploration,it is found that it exists the phenomenon of agglomeration among the nodes of network.It is characterized by the relatively close connection between the points in the same group,while the connections between the groups are loose.The analysis of this phenomenon can not only help us more clearly understand the formation mode of the real network and reveal its meaning,but also help us to understand the functions of the network and make a correct judgment of the event.Over the past few years,a number of strategies and algorithms for exploring communities have emerged.However,these algorithms rarely achieve satisfactory results on both stability and effect.Aiming at these problems,in this paper,it is considered that in the unweighted and undirected networks,the edges of which can only reflect the relationship between two nodes and the differences between these edges are not indicated.It is one of the reasons that accurate results cannot be obtained by dividing them based on the original topology structure.Whether in social networks,transportation networks,or communication networks,the connections between nodes should be different.According to the "echo chamber effect",this paper weighted several powerless networks,verified and analyzed the results by using modularity strength as the evaluation index,and found that there is a phenomenon that the intra-community connections are stronger than the inter-community connections in social networks.In the animal network,the weight of intercommunity edge is higher than that of intra-community edge.Under the inspiration of an algorithm called differential evolution algorithm,this paper proposes a reverse SHADE algorithm,an improved DE algorithm,combined with modularity strength.Compared with SHADE,its performance on global searching has been improved by adding two parameters to control the evolution direction of particles.Benchmark test results proved it.Ultimately,this paper applies the algorithm proposed before to community detection,divides the artificial networks and real network into communities,and makes a comparative analysis with SHADE and IMADECD.It is shown that the algorithm we propose in some degree has improved on computing stability,detection efficiency and accuracy.