Research On Analysis Of Information Diffusion Characteristics And Prediction Methods In Social Networks

With the development of information technology,online social network has become one of the most important ways for people to express their opinions and obtain information in today's society,and the dissemination of information in social network will have a significant impact on the country and society.Therefore,it is of great significance to study the communication rules of information in large-scale social networks for social stability maintenance,public opinion control and information publicity.Existing studies on the rules of information diffusion mainly focus on the dynamics of communication in complex networks,and propose a universal communication model to summarize the rules of information diffusion,while ignoring the impact of the unique social relations and attribute characteristics of social networks on information diffusion.In order to solve this problem,this article first to the law of the social network information diffusion characteristics into the analysis,based on the basic characteristics of social networks to improve the independent cascade model,put forward the calculation method of information diffusion probability,build social network information diffusion model to simulate the actual information dissemination process,to evaluate the propagation influence of social network,On this basis,we further study the prediction of information diffusion in social networks.The main contributions of this thesis are as follows:1.This thesis studies the construction of information diffusion model based on social network characteristics.Most of the existing information dissemination models are constructed based on statistical laws of information dissemination,which are universal but lack pertinence and difficult to accurately simulate the dissemination process of actual information in social networks.To solve this problem,this thesis proposes an information diffusion model in social networks by mining the characteristics related to information diffusion in social networks,and gives the probability of information diffusion to simulate the process of actual information diffusion.Experiments show that the information diffusion model constructed in this thesis can more accurately simulate the actual information diffusion process in social networks.On this basis,this thesis carries out several communication experiments to evaluate users' communication influence in social networks,and the experimental results show that suppressing the nodes of high communication influence in the process of communication can reduce the spread scope of information in social networks to a greater extent.2.This thesis studies the prediction method of multi-feature fusion in social network information diffusion.In this thesis,users' information dissemination behavior is firstly expressed,users' interest in information in social networks is calculated,and multiple characteristics such as influence,interest,attribute and trust of social network users are further integrated.Neural network is used for training,and a prediction model of user information dissemination in social networks is constructed.Attention mechanism is used to verify the influence of trust characteristics on social network users' information dissemination behavior.The experimental results show that the prediction model of user information dissemination has 70% training accuracy for 88% user data participating in training,and 85% training accuracy for 40% user data participating in training.In social networks,the higher the accumulated interest of social network users in information,the larger the scope of information dissemination,and the process of information dissemination will be different because of the spread influence of the initial user node.To sum up,this thesis studies the process of information diffusion in social networks and features extraction,and puts forward the problem of communication prediction based on the analysis of interest and communication behavior.