Construction Of The Complex Network Model And Its Application On Knowledge System

Complex networks can not only describe the complex self-organized system, but also can be used to model and analyze it. If the "element" and "relationship" are abstracted by the node and edge respectively, we can study text knowledge mining from the point view of complex networks. Because knowledge management is very complexity and interdisciplinary, which combining technology and litterae humaniores, we studied the knowledge system integrally from the point of system science and system engineering. Because knowledge is very abstract and there are complex relationship between the owner and carrier of knowledge, knowledge system is considered as a complex self-organized system. In this paper, the evolving mechanism and some dynamic characteristics from the point of system structure determine system functions and system functions influence system structure is studied. Using the complex network theory, a micro-knowledge system of an enterprise is presented. The macroscopic development tendency, middle code system adjustment and microscopic field knowledge evolution using complex network theory is studied.Firstly, the evolution mechanism, which can produce different assortative coefficient r, of complex networks from the point view of topology and weighted are studied. The multistage random growing scale-free network model, growing scale-free networks with tunable assortative coefficient and self-learning mutual selection weighted network model are presented. Under the exogenous pressure, which is represented by the diameter constraint, we studied scale-free network evolving model. To the directed one, the relationship between in-degree and out-degree is studied and found that the exponent of in-degree distribution is determined by the average out-degree.Secondly, the robustness, synchronization and spreading characteristics of network from the point of "the structure determines the dynamic, while the dynamic has some function on structure" are studied. Using the optimization algorithm, we optimized one dynamic character and found the relationship between the dynamic and network structure. The research work of opinion spreading and consensus formation on square lattice have found that the boundary is small than the traditional model.Finally, the complex network theory is applied to an enterprise microscopic knowledge system. By constructing topology and weighted networks (RAN), the macroscopic development trend and the hotspot fields from 1999 to 2004 are analyzed. Based on the text content, the content network is constructed and the classification and clustering system characteristics is studied. Using a faster clustering algorithm presented by ourselves, one can find the community structure in each year, which may give a theoretical foundation to adjust the research code system. The node is defined as one research field, which is presented by one code. If there is at least one proposal filled two codes, there is one edge between these two fields. Using this definition, one can find that, the average distance decreases D with time, while the clustering coefficient C increases with time. In the weighted networks, we presented the definition of "node fitness", which means the nutnber of proposals which have filled only one code. The analysis of the weighted network indicates that the local clustering scales as C_i～k^-1, which has been found in hierarchical networks. Considering the text content, we constructed the content network. We use the vector space model (VSM) to express the text content, and define the text similarity using the cosine definition. The research on the content network may be helpful to study the real development tendency from the content point of view. In addition, a new clustering algorithm is developed for the weighted network and applied it to RAN to find the community structure in each year. Furthermore, whether the community is reasonable or not is studied. These works may be helpful to understand the development trend of Chinese nature science basic research and to adjust the code system.