Construction And Application Of User Influence Diffusion Model In Sina Weibo Social Network

With the development of online social network,the user groups of social platforms are growing.Sina Weibo platform has more than 400 million registered users and more than 100 million active users.Sina Weibo is the largest microblog social platform in China.User share information by publishing microblogs.Other users can make feedback on information by commenting and reposting.There are many applications similar to marketing in Sina Weibo platform due to the huge numbers of users.How to effectively carry out marketing and how to make use of characteristics of social network under Sina Weibo is a valuable research problem.Based on the user influence diffusion model of social platform,the marketing effect can be optimized by utilizing the influence of users,which has practical value.In this work,a social network user influence diffusion model is built to study the influence relationship between users and the diffusion of user influence.The main contributions of this work are as follows:1.The Inf-embedding model of user influence diffusion is proposed.The model includes three main parts: data preprocessing,influence probability calculation and influence diffusion.In the data preprocessing,the maximum likelihood estimation is used to deal with the problem of missing user data in Sina Weibo comments.In the influence probability calculation,a user representation learning algorithm is proposed according to the characteristics of Sina Weibo social network.Reposting and commenting are used to capture the influence relationship between users,which makes the influence relationship between users can be more accurately estimated and predicted.In the influence diffusion,this work takes the independent cascade model as the basic model,combines with the characteristics of Sina Weibo information diffusion,simulates the dynamic process of user influence diffusion,and finally predicts the result of user influence diffusion.2.Based on the real dataset of Sina Weibo,this work verifies the performance of Inf-embedding model from two aspects of influence relationship detection and influence diffusion.In this work,Inf-embedding model is compared and evaluated with existing DTIC-Jaccard model and DTICCredit model in detail.The proposed model is superior to existing models in terms of precision,recall,F1 and other indicators.It shows that Inf-embedding is more suitable for the estimation of Sina Weibo social network user influence diffusion.3.Design and implement a marketing system based on user influence.Users can get marketing suggestions about the social network by uploading the related data.System functions include data upload and parameter setting,social network construction and visualization,user influence visualization,marketing suggestions,marketing result prediction,etc.In the social network construction and visualization,the data preprocessing of Inf-embedding model is called,so that the system can build a social network in the case of missing data.In the visualization of user influence and the prediction of marketing results,the influence diffusion of Inf-embedding model is invoked,so that it can simulate the expected result of marketing and the influence of users through the model.With the system,users carry out more reliable and efficient marketing.