Research On Information Dissemination In Social Nework Based On Interactive Behavior Characteristics

With the rapid spread of Web2.0 technology, interactive services based social network has been an unprecedented development. By sharing, communication and exchange interactions, social network has become the major platforms of information dissemination or rumors spread. Compared with the traditional information dissemination, the time distribution characteristics and relationship strength of the interaction on social network make information dissemination or rumors spread appear more complexity and uncertainty. In addition, overly simplified model assumptions and simple theoretical methods in the traditional information dissemination model can not accurately portray the dynamic evolution mechanism of information dissemiantion or rumors spread process.Based on that, this paper uses human behavior dynamics, complex networks and complex systems theory, transmission dynamics, probability theory and mathematical statistics, and other disciplines for refenrence. Starting from the features of interactive behavior, we build mathematical model for information dissemination and rumors spread on social network. The results of mathematical modeling are basically the same as the results of empirical analysis in this thesis.This thesis has a high theoretical and practical value. The main contents and innovations of this thesis include the following four points:(1) Study the time distribution of interactive behavior on the social network. In the relationship between the rate of behavior execution and the rate of behavior arrival to the research perspective, building convection-diffusion equation, using queuing theory based on priority, we strictly proved that the time distribution of interactive behavior can be portray by power-law distribution of the power of 1.5 when the rate of behavior execution is less than the rate of behavior arrival. Moreover, we also strictly proved that the time distribution of interactive behavior presents polymorphic when the rate of behavior execution is higher than the rate of behavior arrival. The polymorphism is that the time distribution of interactive behavior can be respectively portray by power-law distribution of the power of 1.5 or the power of 2.5 which has an exponential effect when time is less than τ0 or higher than τ0. Whether actual data test results or system simulation results have indicated that the conclusion of theoretical analysis is reasonable.(2) Study the time distribution of interactive behavior on the social network how to affect information dissemination. This paper introduces the propagation tree model based on the infectious disease mechanism of SI model and the generating function method to strictly prove that the changes of node influence has polymorphism on the information dissemination under the condition that the active time of nodes follow gamma probability distribution. The polymorphism is that the number of infected nodes follows exponential distribution when propagation time t<τ0, while t>τ0 it follows polynomial distribution.(3) Study the relationship strength of interactive behavior how to influence the effect of the viewpoint adopted. We introduced a new method which based on the relationship strength theory, which was put forward by Granovetter, to identify the relationship strength between two nodes connected. In the simulation experiments about the adoption of viewpoint, we used the single source of infection with Susceptible-Infected-Recovered (SIR) model, assumed the infection rate is λ=βa, to analysis how relationship strength affects the adoption of viewpoint. The results find that strengthening the authority of viewpoint and friends recommended strategies are conducive to enhancing the probability and velocity of the adoption of viewpoint. Conclusions have a certain practical value to enhance advertising effectiveness on aggregation groups, accelerate information diffusion and inhibit the spread of rumors.(4) Study on the social network how to rapidly and accurately find out the strong high-impact nodes which strongly affect the effect of information transmission, or spread rumors. In this paper, we introduce the model of conductance eigenvector centrality based on conductance between nodes to identify the strong high-impact nodes. The importance of a node depends on the conductance between the node and any other node in complex networks. In random walk, based on the conductance between the nodes, the CEC value of a node is assigned. Steady CEC value of each node is acquired by iterative calculation. Based on simulations of the single source of infection with Susceptible-Infected-Recovered (SIR) model, experiments on sample date set and more standard data sets found the CEC algorithm performance of identifying influential spreaders is superior than the traditional approach of the K-degree, K-shell, RWR and PageRank. Compared with the PageRank algorithm, the difference analysis of experiment found CEC algorithm is more suitable for the average clustering coefficient larger complex networks.