Measurement And Analysis Of A Large-Scale Online Social Network

It’s long been said that the revolutions in communications and information technologyhad given birth to a virtual world. Now it comes true, and our software, computers and cellphones have become woven into every aspect of our lives. OSNs make the cyberspace realand uses depend on it every single day. Most web sites put Facebook and twitter icons ontheir home pages, which makes the World Wide Web more social and more real-time.OSN becomes an important part of our lives, which attract many users and build a virtualworld, which is a map of the real world. Users of such systems will continue grow, andthese systems will change the real world, for they are good communication applicationsafter Email and Instant Messenger(IM), which have made us more interconnected than atany time in human history. Thus, OSN have received significant attention from bothindustry and academia. More and more researchers pay attention to OSN, and manypapers have been published up to now.In microblog system, users connect with each other and form the overlay of the system;users generate tweets and share them with their followers; then, the tweets are spread overthe overlay. Hence, there are three basic questions in microblog system:(i) How do usersconnect with each other and form the network;(ii) what do users talk about in microblog;and (iii) how do the tweets transmit through the network. The goal of this paper is only tostudy how do users connect with each other and form the network using standard complexnetwork techniques. Topic detection and hot topic prediction are very important researchcontent in online social network. However, they are beyond the scope of this paper.This paper spends about two years to observe the overlay of Weibo persistently whichis less studied before, and to make the characteristics of Weibo s network clear, which isthe basic research before researchers could talk about the tweets diffusion and hot topicdetection. All users in Weibo are Chinese and they enjoy a different culture. Meanwhile, itis a map of the real social of China, and understanding its overlay characteristics couldmake us infer the real structure of Chinese society and estimate the influence to the realworld. Our data shows that Weibo has a core/periphery structure, which makes it a goodsystem for information sharing. Then we give one general model and one detail model todescribe Weibo s structure. For the imbalanced development of economy of China, mostusers in Weibo live in the developed cities of the developed provinces, and their activetime is from6AM to24PM. Less than0.03%users draw about30%following links inWeibo, which is not analyzed quantitatively in Twitter, and this is a novel feather ofWeibo. Compared with Twitter, the reciprocal rates of users in Weibo are lower. Thispaper also first finds users with more followers and their followings often follow eachother, meanwhile, they like to post more tweets. The distribution of users followingsand followers partly fits to the power-law distribution. The overlay of Weibo is dynamicalbecause of new users join and existing users change. The reciprocal rate of users islower than Twitter s. This paper proves the CPL is very short and this theory could bewidely used in OSN. We also show the characteristics of sub-groups in Weibo and presenta method to rank the importance of users in Weibo. In addition, we present a methodnamed Friendfinder to detect communities in Weibo, and present a technique namedWeiRank to rank users in Weibo.In conclusion, this paper makes insight into Weibo, and show the characteristics ofWeibo s overlay systematically and comprehensively. As far as we know, this is the firstquantitative study on Weibo s overlay particularly. Understanding the characteristics ofWeibo s overlay is the basic research before other researchers talk about the diffusion oftweets, hot topic prediction and so on, and it also uncovers the characteristics of the realChinese society. Our results are helpful for OSN operators and other researchers on OSNs.Finally, hot topic detection and diffusion of tweets are the two areas for our future workwhich are based on the structure of Weibo, and such researches could make all thecharacteristics of Weibo come to light drastically.