Research On Information Diffusion Model Constrained By Risk Factor

The high autonomy and openness of online social networks(OSNs)change people's way of life,but OSNs will also bring some negative effects.For example,the wide spread of public opinions information of some natural disasters on OSNs may cause panic.Therefore,the study of information diffusion mechanism on social networks can help the government or enterprises release positive information more effectively.Many studies assume that information dissemination is on a social network without weights,so the probability of information dissemination can only be set as a fixed probability of propagation,without considering simultaneous interpreting of different communication risks.In addition,the existing information diffusion models do not consider the types of nodes in the social network that have the exposed state.The exposed nodes carry information in social networks,but they don't spread information.In the paper,a information diffusion model constrained by risk factor is studied.The main contents are as follows:Firstly,a method based on attention mechanism is proposed,and the method can capture the features of nodes,the method can obtain the risk factor of information diffusion between nodes.The method can calculate the attention vector of each node via the features of each node on social network.It is considered that the probability of information diffusion between nodes is higher in the same topic community.Therefore,the attention of node is used to calculate the risk factor for information diffusion.And the risk factor is regarded as the risk factor of information diffusion.Secondly,the paper studies an information diffusion model constrained by risk factor.In the model,nodes are divided into four states: susceptible state,exposed state,infected state and recovery state.Based on the existing SIR model,the exposed state node is joined,and the birth rate and death rate of the node are increased.In order to better express the difficulty of nodes,the risk factor are applied to the diffusion model as the infection rate of information diffusion.At the same time,the influence of different infected rate,conversion rate and recovery rate on the diffusion model is studied.Finally,the experiments were conducted on three datasets: Facebook,Facebook sub and Cora,verifying the feasibility of the above model and method.