Mining And Modeling The Features Of Social Networks Influence Based On Sina Microblog

In recent years,the online social networks have become unprecedented popular.The rapidly developing online social networks have played a vital role in human life.This have made social influence the most written topics.Previous researches have used the features like topological structure,user popularity,user behavior and so on.Some research even proved the relationship between topics and influences.But most studies lack of analysis between topics,users and diffusion patterns and they haven't fully consider the relation between user actions and influence.Actually,there are lots of applications for users to check their influence in social networks,but the combination with the real application field is not quite enough.The thesis is based on the research about social influence of the predecessors.In the first part,the thesis analyzes the behavior of sina users to verify its characteristics of complex networks.Then classify the diffusion patterns,users and topics,explore their relevance by using FP-growth methods,mining different topics,different users with different diffusion patterns.In the second place,through the shuffle test and reversal test to explore the causal relationship between the user retweet behavior and their friends behaviors,and then analysis respectively on different topics,explore the causal relationship between user retweet behavior and user interests.Eventually proved that in sina microblog the relation between user retweet behavior and the number of retweet friends is not obvious.It is the user interest and friends retweet behaviors mutual results.Then,mining and define three characters features including the verified type,in-degree,and activeness user.The above experimental results show that in sina microblog,the influence of the users is not only decided by their friend relationships,but also by the correlation between user and topics,audience and topics.Therefore,based on the PageRank algorithm,the thesis consider the correlation between topics and posters,audience relationship factors and users' character features design the influence model based on topics,and verify its effectiveness through comparison with 4 common models.At the end,the thesis uses the social influences of users and diffusion patterns to design and realize a microblog marketing delivers finding application.According to the topic of marketing,the application combine with the influence ranking model algorithm,and recommend different diffusion patterns to the advertisers.The application accomplish apply of the influence rank and satisfies the advertisers' individual needs.