User Influence Analysis In Micro-Blog

With the rapid development of internet and technology, people can share and exchange information more conveniently. The speed and scale of information transformation has attained unprecedented levels, especially after the emergence of micro-blog.Micro-blog is a broadcast social network platform which uses "following" mechanism to share brief real-time information. It is the hottest social media because of four characteristics-anyone, anytime, anywhere, and anything. Micro-blog marketing arises at the historic moment, and one of its important requirements is to maximize user influence. Therefore, to select the most influential user in microblog-sphere is becoming one of the most important research areas. The research mainly analyzes the influence of a user’s action on other users’ action in a micro-blog community. Since resources are usually limited, information can be retrieved and spread more efficiently by the most influential users.Influence has been detailed studied in the field of sociology, but the majority of the study is qualitative analysis, while there is only a little quantitative analysis. In this thesis, user influence in micro-blog is quantized based on the previous work, then detailed analysis and comparison are carried out.This thesis focuses on user influence analysis in micro-blog:There are some users, each user has his/her own followers, friends, and messages etc. Our work is to select the most influential user from microblog-sphere.We first analyze the network structure of users to get the following algorithms: the Follower-Friend algorithm, the expected influence of K-coverage algorithm, and the Average of Posts and MicroRank algorithm. After doing analysis and comparison on the performances of those algorithms, we found:1. There is no contribution from the users who has more than5indirect attention to her or his friends.2. The number of followers and friends only reflect the expected user influence, there are some other factors that may be related to the actual user influence. Then we put forward two methods for studying user influence taking user interactions into account:K-coverage algorithm and Microblog algorithm. User interactions include for example commenting and forwarding friends’messages. At last, we compare the performance of the proposed six algorithms on the sina micro-blog data by making Kandall’s correlation coefficient analysis. We found that Microblog algorithm has the hightest correlation with the Average of Posts algorithm, that is, Microblog algorithm is closer to the effect of forwarding messages. The results show that considering user interaction to user influence is worthwhile.