Research On Local And Overlapping Community Detection Of Weighted Network Based On Community Stregth

Complex networks are ubiquitous in the real world,and the research on complex networks has penetrated into every subject area.Community structure is one of the most important characteristics of complex networks,and we can get a lot of useful information in real networks by studying the structure of community.At present,there are a lot of algorithms for the community detection in complex networks,such as graph partitioning method,hierarchical clustering method,modularity-based method and the method based on statistical inference.These existing algorithms are mostly for unweighted networks.However,many real network is essentially a weighted network,weights of the edges has a important influence on network's function and performance.If we ignore the weight information,a lot of useful information will be lost.On the one hand,with the reality of the network continues to expand the scale,topology structure is becoming more and more complex,which making us difficult to grasp the information and structure of the whole network.Hence,we need effective local community detecting methods to solve the problem.On the other hand,some nodes are in a number of communities in the actual network,which are called overlapping nodes by researchers,and the corresponding communities are called overlapping communities.At present,there are a number of overlapping community detection methods,but most of them are global algorithm and mainly used for the detection of unauthorized network,not suitable for detection overlapping community structure in such weighted networks,which are large scale,and have complex structure without knowing the global information.Therefore,this paper will do research on the local community detection and overlapping community detection of weighted network.For the local community detection,we propose a simple,fast and effective method,define the community strength coefficient,extend the p-strength measurement to weighted networks,verify on the computer generated networks and real networks,and compare with some of the existing methods.The results show that our algorithm has lower computational complexity,can accurately detect network's community structure.For the overlapping community detection,based on the local community detection method of weighted network,the idea of community strength and modularity optimization,we can detect overlapping nodes accurately,and then using the method of normalization to calculate the fuzzy coefficients of overlapping nodes.Finally,we carry on the experimental analysis to the computer simulation networks and real networks,to verify the validity and accuracy of our algorithm,showing our algorithm overcomes the shortcomings of existing overlapping community detection methods which are mostly aiming at the unweighted network of global information.In the first chapter,the research background,significance and current situation of the complex network community detection are studied,and the research contents,innovation points and the structure of the thesis are introduced.In the second chapter,the topological characteristics of complex networks and the commonly used models,the definition of community and the measurement of community detection algorithms are studied.In the third chapter,for weighted network,a local community detection method is proposed based on the community strength coefficient.In the forth chapter,combined with the characteristics of the local algorithm,a local overlapping community detection method is proposed on weighted network,and can calculate the fuzzy coefficient of overlapping nodes.Doing experimental analysis on simulation and real networks,and comparing withthe existing algorithms to verify the performance of our method in the third and forth character.At last,we summarize the research results of this paper,and discuss the prospects.