Based Microblogging Social Network Features

Complex science is the science of21st century, which aims to unravel the dynamic phenomenon that cannot be explained by the current knowledge. In this thesis we applied the complex theory onto social analysis to reveal the diverse phenomenon, culture as well as social behaviors of human, especially group behaviors. Studies have shown that online social communications as blogs, BBS and chatting rooms etc. are the close reflections of relationships people and organizations in real life, and thus the data generated are suitable for analyzing the patterns between individual and group, the interactions of groups, the morphology of social groups and ways of which they evolve.This thesis studies the Microblog network, use complex network theory, system dynamic theory to explore the features of online instant messaging social network. In this thesis, we will investigate the power law properties in human behaviors and social systems from the empirical data, the modeling studies for the underlying mechanisms, and the effects on social dynamics.The main works concluded are as follows:(1) Investigated the dynamics properties in human behaviors and social systems from the empirical data. The human behaviors shows "long period silence and short period high frequency bursts" in both time and geographical behavior patterns. The patterns follow the fat tail distribution. Especially in the time behavior pattern we obsevered segmental power law distribution, which follows the characteristic of human society. We modified the BA model and implied Markov Link theory to model the link-in probability of a node in Micro-blog network. The out-degree of nodes in Micro-blog network also show segmental characteristic, which may resulted in the existence of "star node" and recommend mechanism. The results also demonstrate that Micro-blog network has short node path length and high cluster coefficient.(2) Next we inspected the community structure in Micro-blog network. We defined the network community topology with closed edges and inside tightness and give the definition and restraints. Then we proposed a subgraph and structure isomorphism based algorithm:Frequent Subgraph Mining (FSM). The algorithm stores the frequent subsets and graph sets by decision tree, and save every permutation matrix of graph in graph sets. By this we guarantee the complicity of frequent subgraphs and its topology. Then, algorithm traverse the candidate subgraph sets and disposes the improper ones. The final time complexity isO(N*M*(I+N)). The algorithm reduces search range and improves system performance.(3)Finally we studied the information diffusion process in Microblog network. By empirical data collection we proposed a five-stage evolution model for public opinion on emergencies. The model is from the first stage (lowest)"social silence or chatting room mode" level to the fifth (highest)"body conflict" level. We find that a concept can be diffused much quickly on Microblog network than in the real world, and withered away quickly as well. Study shows that the wither-away process can be so quick that it can drop from the fifth stage directly to the first stage. Then the diffusion model of concept is studied and we proposed a mean-field based diffusion model in Microblog network, the model is based on BA model and under the mean-field assumption and we divide users into three groups:"non-enlightened" group,"enlightened and committed" group and "enlightened and non-committed" group. Experiments show that the lager the "enlightened and non-committed" group the more rational the network would be. And the transmission ways is insignificant referring the result. Finally the "Weak Tie" theory is analyzed on empirical data.