Research On Community Dynamic Evolution In Complex Networks Based On Local Betweenness

The rapid development of network technology brings us a lot of resources, it becomes an important research topic that how to find out what the people need quickly and efficiently from the vast network resource today. The study of scholars indicated that these data are not unrelated, on the contrary, there are some links among the resources, which exist in the form of community structure in the network. As community structure is an important feature of complex networks, so the study of community structure is not only conducive to a better understanding for complex networks, but also there are important theoretical and practical significance to solve difficult problems in the real world.Firstly, this thesis measured the properties of complex network, studied several typical network models of complex network, and gave a brief description of social networks. Then we analyzed these algorithms which exist in social networks, that some algorithms have higher accuracy in discovering community at particular networks, but have low time complexity; while some algorithms have better effect in dealing with the chaotic structure of the network, but they are only applicable in some networks which have particular network structures. Lastly, based on analyze the dynamic structure of complex networks, the algorithms of discovering non-overlapping communities and the algorithm of discovering overlapping communities in dynamic network proposed after a series of studies.The dynamic characteristics of complex networks make the community detect that time axis needs to be added in algorithm in order to make the results of internet community more real. This thesis puts forward the idea of finding non-overlapping community algorithm in dynamic network, which is based on the ideological basis of GN arithmetic betweenness, at the same time, taking the localization into consideration during the process of calculating the network edge betweenness largely improves the time complexity of calculation. First of all, algorithm integrates the dynamic network information on time axis, and attenuates the historic information based on the sociology experience, which makes the network data information closer to real-world data information. Secondly, according to the feature that the biggest edge in network side vertex betweenness is the key edge, this thesis removes the biggest edge in network edge vertex betweenness and gets disconnected, which makes a community structure. Then, based on the three sets of data, this thesis conducted the arithmetic testing and analyzing, and found the community divided result was relatively correct. At last, based on the data sets in different order of magnitude, this thesis made a comparison between Framework model and algorithm FacetNet in time complexity contrast, and found that the algorithm in dealing with large data sets have obvious advantages.Based on the facts that a network node may exists in various community, this thesis brings forward a new algorithm of finding overlapping community in dynamic network. This algorithm is based on the foundation that found the non-overlapping community in dynamic network, rethinking the network structural characteristics and get online community structure by copying the biggest node in network segmentation vertex betweenness. The algorithm process is based on the non-overlapping community algorithm, joined the largest node of copy segmentation betweenness, and to get disconnected community found. Finally, analysis the algorithm in C-DBLP data set and global food export trade data sets.The two dynamic algorithms proposed by this thesis, using the ideology of localization, there is a great improvement in algorithm time complexity, and the higher accuracy of the algorithm found community, which is worthy to be popularized and applied in the real world.