Research On Overlapping Community Detection In Complex Networks

Community discovery algorithm in complex network has a very important theoretical and practical significance, therefore, it has been received widespread concern of many scholars, which are from computer science, mathematics, physics, biology, sociology and economics. Their research has obtained important results, such as the GN algorithm which represents the non-overlapping methods and the Clique Percolation Method which represents the overlapping communities detection algorithm that allows the nodes belonging to different communities at the same time. In terms of how to evaluate advantages and disadvantages of the results of community discovery algorithms. Newman et al put forward the concept of modularity, and scholars had expanded modularity to a variety of community networks including overlappping, and got different forms of correspond expression.This paper introduces the research background, significance and research situation of the overlapping communities discovery algorithm in complex network at first. Second, we describe the basic theory of complex network, including the historical progress of complex networks, the network graph model representation, and a variety of structural properties of complex networks, as well as a variety of complex network model of network topology. The third, we discuss the concept of community, and in-depth research the community detection evaluation criteria--modularity. We also in-depth research the current community discovery algorithms, in particular the overlapping ones, we find that there are many shortcomings in current algorithms. Such as, the overlapping detection algorithm LFK, it can not guarantee the same result of multiple runs, and it also easily leads to big overlap degree although it can cover the whole network.For the shortcomings of the current algorithms, we learn from the idea of natural communities, and modify the criteria of determing the formation, get our algorithm for natural communities. We find that many factors influence community overlap, after we study it. The communities neighbors overlap of overlapping communities occupy a certain proportion of judgments, so we summarize the standards of community overlap, and give the conpet of Community Overlapping Degree, for short is COD. After forming of natural communities, there is a large overlap of many natural communities. We take into account the actual situations, these natural communities may be wrong divided into multiple communities. So we add the merger stage of natural communities. We take the COD as the standard of merging natural communities, get our overlapping community detection algorithm OCNF, which is based on adjustable fitness and share neighbors. Experimental results show that our algorithm can get good results by adjusting the different parameters. The OCNF can successfully detect overlapping nodes from network, however, there are still some areas for improvement, such as how to reduce the complexity of running time, how to better display the results graphically, and so on. Which still need us to do more work.