Research On Information Propagation Method Based On User Feature

With the rapid development of the Internet,the popularity of handheld terminals and social platform to promote the dissemination of information has undergone a fundamental change.Microblogging as an open platform,the concern between users do not need certification,information dissemination has a fast,wide range of features.How to effectively predict the spread of information has become an urgent problem to be solved.At present,the research of information dissemination model based on social social platform mainly focuses on the influence of social network structure and information content on information dissemination.And the impact of individual characteristics and new users on the impact of information dissemination is not deep enough,and then for the lack of historical information users,proposed by friends and friends to predict the behavior of microblogging forwarding program.For microblogging platform,microblogging forwarding is the basis of information dissemination.This article focuses on the microblogging forwarding will be affected by what the individual characteristics of the user,how to effectively predict the user's forwarding behavior through these features.First of all,this paper regards the forwarding behavior of micro-blog as the interaction between three entities,such as sender,receiver and micro-blog.When extracting the characteristics of the user,consider the individual characteristics of each entity first,and then consider the characteristics of the relationship between the two entities,this paper extracted a total of 11 features.Where the authoritative feature is measured by the PageRank algorithm,which reflects the location characteristics of the sender in the network.For the similarity of interest,the LDA model is used to model the user's history microblogging information,and the distribution of the microblogs is obtained.Then the improved KL algorithm is used to measure the similarity of the two users' interest.The extracted feature is normalized and the SVM algorithm is used to predict the microblogging forwarding.At the same time,the influence of the non-equilibrium data set on the SVM algorithm is also considered.Using the SVM-based oversampling method to optimize the data set can improve the prediction effect.Second,this article also build a friend selection model,for the lack of historical information users,through the friends selected model from the adjacent friends to choose a close friend to predict microblogging forwarding behavior.The principle of choosing a friend is: for the same microblogging,the more similar the forwarding behavior,the more qualified as a close friend.The method proposed in this paper is verified by the real data set of Sina microblogging,and compared with the logistic regression algorithm and Naive Bayesian algorithm.The experimental results show that the proposed method can improve the accuracy of the prediction results effectively,and the algorithm proposed in this paper still has a good effect when the other algorithms are invalid for the new users lacking historical information.