Research On Community Detection Algorithms In Complex Networks

Networks have a broad definition,as long as there is a connection or interaction between any objects will form a network structureand network is universal in real life.We can grasp the laws that exist in the network by researching network and mining the useful information.Finding the community structure through community detection algorithm is the premise of in-depth study of networks characteristics,and it is good for solving practical problems.In this paper,we present the integrity of community and present a new evaluation function F base on the concept of community integrity.First,we prove the boundedness of evaluation function F.Then compare classical modularity Q with evaluation function F,the results show that evaluation function F can evaluate the detection quality of community reasonably and the sensitivity of evaluation function F is better than modularity Q.The essence of community detection algorithm is dividing network into locally dense sub-networks.According to this view,we present a new non overlapping community detection algorithm base on community density(BDA).The core of the algorithm is selecting the initial community which has the maximum density and extending this community.Verifying BDA algorithm on real data sets and compare with FN algorithm and LPA algorithm,the results show that BDA algorithm can find the reasonable community structure of small particle size,and new algorithm is better than FN algorithm in the aspect of time complexity with the increase of number of node and edge.In order to meet the actual demand,we present the community merge algorithm base on similarity,this algorithm can merge small communities into large communities.Because overlapping community structure is difficult to determine,we present a new idea that we can extend non overlapping community structure to overlapping community structure and present the specific algorithm base on non overlapping community structure(EOA).The result show that new algorithm can find the reasonable overlapping community structure and new algorithm is better than LMF algorithm in the aspect of time complexity.Finally applying BDA and EOA in the scientific partner network and getting better community structure.