Research Of Supernetwork-Based Internet Social Network Analysis Model

With the rapid rise of Internet technology and applications, many kinds of network application services which offered by the Internet become the important transmitting vector of public opinion. More and more people are using these application services as a main communication method. Social network analysis can be used to model society entities and the relationships in the society. The in-depth analysis of the network entities on these internet services'platform would be an effective way to understand the public opinion and grasp social network's ideological dynamically. The surge of social network analysis based on these Internet services become one of the new hot research topic within the traditional social network analysis.The major contribution of this dissertation is that on the one hand it first applies supernetwork theory to solve the real Internet social network analysis, especially on the problem of social network entity relationship analysis and characterization model and the problem of network structure mining analysis. it also gives a clear research train of thought about the social network analysis on the Internet. We mainly focus on the problems of data mining from the Internet, characterizing the relationship within the internet entities, and community detection. The main results are as follows:1) Data mining huge amounts of information from the Internet, this thesis proposed the information fusion technology of extensible information grabbing and presenting technology. We develop an information acquirement system for data mining from the Internet which can be used in automatic text categorization and standardized storage technology.2) Research of characterizing the relationship of network entities. For the problem of existing relation representation methods which are based on single factor cannot truly characterize the actual relationship within the internet entities, the dissertation put forward a supernetwork- based model to characterize the implicit relationship of web entities corresponding to their interests, recombination with dominant social relationships. This is a more realistic multi-dimensional relationship model to characterize the Internet entities relations. The output of the model is a weighted relation network which would help for predicting the attention scale of potential hot topics in Internet.3) Mining potential network structure. Using the existing research of mature community structure detection for reference, according to the low resolution ratio problem caused by the modularity function algorithms which are used widespread, considering the entity active degrees, the thesis proposed a novel community mining algorithm combining the pretreatment method of utilizing the tie strength information with optimal method of the modularity function hierarchical agglomerative clustering. This new algorithm can find smaller communities which have more meaningful connecting information than the traditional algorithm. By filtering the index on the entities'activity also reduces the computing dimension. Meanwhile, the algorithm's complexity is close to linear time, which can be suitable for analyzing the real large-scale network structure. The outcome of the algorithm has a certain value for describing the network structure of internet and predicting the tendency of network public opinion.