Analysing The Topological Structure And The Content Relavence Of The Information Networks

The popularization of Internet promotes the development of the networks in social information. It is necessary for analysing the significant characteristics of diverse large scale information networks. There are so many challenging problems, such as how to model the diverse and complex information networks, how to get high efficient results in large network information retrieval, how to mining deeper content from the networks, what's the essential rules of topic promulgation in information networks and how to predict the spreading behavior in the networks etc. After analysing the traditional topological characteristics of the Web, we find that there exists a kind of high coherent relationship between the topological clustering and the content clustering in the information network based on microview. Based on this kind of relations between the topological structures and the content distributions we study the web modelling, community identification and some related application problems in detail:First, after some existed characteristics of the Web topology are verified, some new characteristics are discovered : the high clustering property in micro-topology (high average gathering coefficient), the obvious mapping relation between the topological struture and the content in micro-level,linear irrelevant between the degree distribution of network nodes and the relative degree distribution of contents etc.Then after analysis the topology of the complex network and the network modeling, the muti-scale determinism is proposed, especially for the information network a web evolvement model(PRCP Model) that fused the node authority and the node correlation is proposed. The model deduction, evolving learning verification and large scale experiment proof indicate that the model can explain the micro-topology centralizing phenomena, can imitate the mapping relation between the network connecting distribution and network content relative distribution and also can predict the mapping relation between the topology clustering and content clustering.