Modeling Of Information Diffusion And Popularity Prediction In Online Social Networks

With the rapid development of Internet technology,the application scale of online social networks as a new communication medium is increasing day by day.People are not only consumers of information,but also creators and communicators in online social networks,which brings information explosion.In this pro cess,social networks have gradually replaced the dominant position of traditional media,influencing people's thinking and impacting people's outlook on life,values and world outlook.At the same time,the information presents a nuclear fission-type diffusion method,which accelerates the formation of social hot spots,and harmful public opinion has a negative impact on national security and social stability.Therefore,the research on information diffusion of online social networks has important applicat ion value and practical significance.At present,there has been a lot of work on information diffusion research on social networks.However,the study of the mechanism of communication evolution involves multiple disciplines such as network science,socio logy and psychology,and there is still a lack of research results in interdisciplinary subjects.This article revolves around the three elements of "network structure","user group" and "information content" of information dissemination,using sociological theory for reference to study the influence of network structure on information diffusion.In view of the scale-free characteristics of the popularity distribution of information content,the issue focused on the problem of popularity prediction,and in-depth research is carried out on the the datasets of Facebook and Tencent News.The main contributions and innovations of this article are summarized as follows:(1)In the aspect of information spreading modeling,aiming at the problem that traditional work does not consider the impact of multiple redundant contacts of information on individual behavior,the Mainstream Fatigue Theory in sociology is introduced,and a new information spreading model is proposed.By dividing the users in the social networks into four evolutionary states at the micro level,the interactive Markov chain method is used to probabilistically represent the individual's state transition at the micro level from the time and space characteristics of the information propagation,and derive into dynamics equations.The Monte Carlo simulation of artificial networks and real social networks shows that the network structure does have an impact on the information sprading,and if the same message spreads over and over again,which may have a negative impact on multiple redundant contacts of network individuals,thereby reducing the effect of information spreading.(2)In the aspect of popularity prediction,in view of the problems of feature selection and prediction accuracy in the modeling and solving process,a prediction model for Facebook's well-known homepage is proposed.By analyzing the share law of the messages published on the homepage history,it is found that if a message can attract more weakly connected users to share in the early time,it will be easier to obtain greater popularity in the future.Based on the above findings,the tie strength feature is extracted,and the popularity of in the is merged to establish a multiple linear regression equation.By comparing with other representative benchmark models on Facebook's real data set(including 1.54 million shares),experiments show that the proposed model is superior to existing methods in predicting the final popularity.(3)In the aspect of the applicability of the prediction model,for the complex multi-information concurrent environment in social networks,a popularity prediction model based on the competitive matrix is proposed.At present,most prediction models are based on the assumption that information is independently spreading.However,the actual situation is that public attention often shifts due to the occurrence of hot events,which leads to inaccurate predictions.In this paper,by studying the comment data set of Tencent News,the entropy method is used to explain the existence of information competition and the limitation of user attention.Aiming at the representation problem of complex and changeable information content.Massive news texts and user comments are expressed and classified in vectors through neural networks and deep learning methods,and a competition matrix is constructed and solved using gradient descent algorithm.Comparative experiments show that the prediction model proposed in this paper is suitable for a multi-information competitive environment and has better prediction performance.(4)Designed and implemented an information transmission analysis and prediction system,which is applied to the actual needs of Internet public opinion analysis and forecast of emergency event propagation trends.The system is mainly composed of four layers: information collection,data storage,predictive analysis,and front-end display.It can achieve data acquisition,analysis,and prediction for typical social platforms such as Facebook,Tencent News,and Sina Weibo,and achieve good practical results.