User Impact Analysis And Research In Social Networks

With the development of the Internet era,all kinds of social network platforms are developing rapidly.The relative subjects of user influence of social networks have become an important research field of the Internet.Most of the existing influence assessment models and the influence maximization algorithms do not take into account the unique user behavior characteristics of social network,also,their time complexity is high and the spread range of influence is unstable.In addition,most of the influence of social networks maximization do not consider the importance of topics for node mining,leading to low coverage of influence on specific topics.In view of the above problems,to mine and analyze the data produced in social networks,study the behavior characteristics of users,combine user behavior characteristics and the influence assessment models together to construct an evaluation model which conforms to the behavior characteristics of social network users,and aiming at the phenomenon of hot topic search in more and more social platforms to search the influence maximization algorithm under specific topic,achieving the effective dissemination of information and control of public opinion.This shows that research on the influence of social network users has very important theoretical and application value.This thesis takes the user influence of social networks as the focus.Firstly,we introduce the related background knowledge and theory of social networks,describe in detail the crucial information propagation models in the dissemination of information,and then analyze the advantages and disadvantages of the existing influence algorithms.Secondly,we combine the user behavior characteristics of social networks with PageRank model,and construct a more accurate and real-time user influence assessment model.Finally,according to the concepts of URIR and Ttop filter user nodes,introducing suitable pretreatment method of social network data,combining K-means and LDA with the theme of mining,we design an influence maximization algorithm based on topic.We use big data analysis to verify the experiment parameters on the accuracy of influence evaluation,the influence dissemination effect,the time complexity of the algorithm,and some other aspects,the simulation experiments are carried out to verify the analysis.The detailed research work of this thesis runs as follows:(1)We propose a real-time influence assessment model that conforms to the user behavior characteristics of social networks,and the model is modeled again by combining the user behavior characteristics in social networks with PageRank model.We discard the method of average distribution of influence weights in PageRank.We redefine the influence distribution factor in order to make the model identify high quality fans effectively.At the same time,we propose the concept of interest and activity to identify zombie powder and active fans.Finally,on the real user data of Sina micro-blog,we use Matlab to verify that the improved model can better reflect the law that user behavior changes with time,what's more,it has a good convergence,and can assess the user influence more accurately.(2)This thesis designs an influence maximization algorithms Topic_MIA based on topic.firstly We use Topic MIA algorithm to mining nodes for the first round of screening.and then select top-ranking users witch under the double standards of URIR value and Ttop value and add them to mining nodes,introduce data pretreatment method under the specific topics,use method of KM_LDA which combined K-means and LDA to cluster and mine theme,and the Independent Cascade Model is modeled again to form the new propagation model called Topic IC,and finally we verify that the improved algorithm on the real user data of Sina Micro-blog.It is more stable and the time consumption is lower by analyzing the experimental results.