Analysis And Prediction Of Information Forwarding Behavior Of Microblog Users

In recent years,with the development and popularization of Internet,social tools have been widely promoted,and the scale of users is expanding.Through social networks,users not only can grasp news information immediately,but also can actively participate in discussing topics of interest and share opinions with others.In social media,which has a huge amount of information content and a large user base,the dissemination and forwarding of information has brought about a strong social influence.Therefore,researching the user's forwarding behavior in social networks is of great value in the analysis of public opinion and the study of cyber group events.Taking the Sina Microblog as the research object,based on the analysis of the main problems existing in the forward behavior prediction of social network users,the thesis focuses on the factors affecting the user's forwarding behavior and the user's forwarding behavior prediction model.First,in view of that the existing user forwarding behavior researches are without considering the individual difference of users,one kinds of user interest degree algorithm based on the topic region related factor are proposed.Both algorithms measure the degree of association between the user's location and the Weibo content by defining the topic region related factors.The difference lies in the range of the geographic area to which the Weibo content belongs.Combined with other common features,the effectiveness and advantages of the improved algorithm are verified by SVM classification algorithm.The experiment shows that the interest algorithm based on the topic region related factor is better,and the prediction accuracy is 3.6% higher than that of the original interest degree algorithm.Second,in view of the low prediction accuracy and long computation time of user forwarding in social networks,a user forwarding prediction model based on sparse representation classification algorithm is proposed.Based on the extracted user's basic characteristics,micro-blog's basic characteristics,user activity characteristics,improved user interest characteristics and user intimacy characteristics,the algorithm uses sparse representation classification algorithm to predict micro-blog's user forwarding behavior.Compared with support vector machines,logistic regression,random forest algorithm,the user forwarding prediction model based on sparse representation has higher accuracy and lower time complexity.