Research On Evolution Model And Information Dissemination Mechanism In Online Social Network

With the rapid development of the Internet technology,online social networks have replaced the traditional media, such as television, newspapers, and become the largest platform for information communication in the world. With the change of the way of information transmission, when people enjoy the convenient and fast of expressing their ideas and views, they are also troubled with all kinds of rumors in online social networks. Information dissemination of online social network has a great influence on politics, economy and culture of human society, and eventually even affect national security. Research on the rules of information dissemination of online social network has an important role in management of information and public sentiment, as well as in tracing and early warning of networking event. And it has the realistic significance in a certain degreery to safeguard national security and social stability.Main research contents of this thesis include the following aspects.(1)On the basis of BA scale-free network model, an evolving model based on mechanism of stochastic growth of new edges was proposed, the model also revealed the mechanism of the phenomenon that the degree distribution in double logarithmic coordinate has the characteristic of head bending. Considering that the numeber of the new node’s friends is usually random in the network evolution process, this thesis proposed three evolving models based on the growth of new edges under different conditions: the number of edges obeying Poisson distribution; part of the number edges obeying Poisson distribution;the new node’s selection probability is entirely random. Then the degree distributions of the models were deduced. The result shows that the random of the growth of new edges leads to the phenomena of head bending of the degree distribution. When the degree is larger, the degree distribution still follows the power-law distribution. And the distribution will not follow the power-law distribution if we replace preferential mechanism with random probability distribution. The simulation networks were created based the three evolving models at last, and the experimental results showed that the assumptions were reasonable and the theoretical analyses were correct.(2)According to the characteristics and some problems in current researches of information transmission in online social network, an information diffusion model based on relative weights of users was proposed. Whether the information is propagated efficiently to other nodes depends on the status of both sides of the disseminator and embracer. The function of the mutual influence between users in online social network was defined, the information communication process and propagation path in network were analyzed in this thesis. We discussed the influence on different paths, and put forward an information diffusion model based on relative weights of users and its dynamic models. Simulation and experiments of the model and SIR model were carried out on nine networks. The authority node and the ordinary node were seleced as spreaders, the different influence of them were studied in the experiments. The results showed that the influence of the ordinary node weaker than authority node. There were significant differences in heterogeneous networks but not in homogeneous network. It proved our model can reflect the real characteristics of information transmission on online social network to an extent.(3)According to the characteristics of rumors spreading in online social network, a new rumor spreading model based on SEIR epidemic model is proposed. As rumor spread is different from the spread of infectious diseases, according to whether the people knew the rumor, this thesis divide the people in the whole social network into four categories: ignorants, spreaders, insider and stiflers. In order to describe the different reaction of people when they heard rumors, we introduce different probabilities of four categories and put forward a rumor spreading dynamic model. Furthermore, the stability of solutions for differential equations of the dynamic model is proved, and the influences of the different probabilities on the solutions are also discussed in the thesis. At last, simulation and experiments of the model are carried out on nine networks, and the experimental data verifies the validity of the theory.(4)Aimed at the limitation of traditional ranking algorithms based on the static network statistics, a new ranking algorithm based on information spreading probability was proposed. The data shows that traditional ranking algorithms cannot solve the problem that the influence of the nodes in the network is different with the spreading probability. Considering the effect of the spreading probability on nodes, this thesis analyzed the influence of the spreader on different layers neighbours, and put forward a new ranking algorithm, which considered both network topology and the influence of the spreading probability. Furthermore, an approximate calculating method was proposed, which greatly increased the computation efficiency. At last, simulation and experiments of the model were carried out on nine networks, and the experimental data showed that the stability and integrated scoring of our algorithm were higher than the other algorithms.