| With the evolution of information technology,social websites and micro-blog have been developing rapidly.Compared to traditional social networking,the online social network shows distinct characteristics such as openness,anonymity,cross-regionality,high interactivity and complexity.The research on online social network and its popularity can help various administrative authorities understand the online community structure,identify high popularity users,provide timely,reliable and valuable information for businesses,and can also keep the government informed of the dynamic behavior and the network users' intentions.This dissertation attempts to study and analyze the social network structure of micro-blog,with particular focus on the community partition algorithms,the community extraction algorithms,and the evaluation of micro-blog users' popularities based on micro-blog social network.The main contributions are as follows:1)Sina micro-blog is one of the most popular social networking platforms.Online social network is a dynamic development process,with new users and new interactive relationships generated every day.Then,how to find the rules of the social network and extract the model of network evolution? In this dissertation,by studying the interactive relationships of the network users,we have constructed three kinds of networks and analyzed their features,which are Follow-network,Fan-network and Allnetwork.2)Many groups exist in online social network,which are difficult to track,because of the imperfection of existing community partition algorithms.Thus,it is very important to find a more accurate community partition algorithm by analyzing the structure characteristics of the existing community partition algorithms.In this study,a novel hybrid algorithm based on game theory is proposed by combining the advantages of the FNCA and LPA.The experimental results of the hybrid algorithm are closer to reality and much easier to reveal the relationships between micro-blog users.3)In global community partition algorithm,the topology structure of the whole graph is required as the input.However,it is very challenging to get the whole topology structure.For example,the scale of World-Wide-Web is huge,so it is almost impossible to download the entire network topology.As a user,it may not be necessary to divide the whole network.What users are interested in is likely to be just the community structure of a known small subgraph.For example,in social networks,we want to know a specific person's community,that is,how to determine a particular node's community.To research and analyze the local community structure,a community extraction algorithm based on local information is proposed.Based on three classical methods-R,M and L,and taking into account the connection density between the internal and external user nodes of a community,the W method is proposed in this dissertation.Experiments show that the W method is stable and more suitable in engineering application,compared to other algorithms which show different performances in different data sets.4)In online social networks like Sina micro-blog,each user's popularity is different.It is difficult to evaluate users' popularity simply from their fans.How to accurately evaluate users' influence? Through the analysis of the behavior patterns of Sina users,a new method based on the knowledge of the electromagnetic field to analyze micro-blog users' popularities is presented.The dissertation proposes the evaluation of Sina micro-blog users' real popularities on Sina micro-blog platform,based on the new concept of a specific user's micro-blog flux density.The experiments show that the method is effective and reliable. |
PDF Full Text Request