Page 221 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42

220 Journal of Library Science in China, Vol. 8, 2016

& Information Science in recent years. Knowledge network is the product of knowledge
exchange,and both classic citation network and co-authorship network are dominant
representations for it. Many scholars studied citation network and co-authorship network to explore
the interaction among scholars,collaboration evolution mechanism,knowledge dissemination and
development from the perspective of network evolution.
Library & Information Science has long been concerned about the evolution model of
knowledge network. Early Price Model laid the ideological foundation for BA scale-free network
discovery. As knowledge network has small-world structure and scale-free characteristic,growth
mechanism and preferential attachment mechanism are essential to understanding its formation
and evolution. Citation network can reflect the flow and the inheritance relation of knowledge
clearly. Moreover,temporal and spatial factors and nodes heterogeneity can affect its preferential
attachment mechanism,and function expression of new nodes and edges can help simulate
its growth mechanism. By the appropriate modification,these factors can also be used for co-
authorship network.
Co-authorship network reflects the collaboration relationship among knowledge creators. Based
on co-authored papers,this kind of network has important significance in understanding the pattern
of knowledge exchange and the establishment of interdisciplinary collaboration. Similarity-based
algorithms can help examine the link prediction effect of co-authorship network and effectively
identify the main influence factors on connection,which lays the foundation for the construction
of evolution model. Most current researches still use BA Model and its improvements to study
evolution mechanism. A future important research direction would be adding some external
attributes of nodes to build hybrid preferential attachment model. As scientific collaboration
contains knowledge exchange process,analysis ideas based on scientific research team evolution
or network co-evolution can help identify the dynamic factors of the evolution and discover the
scientific knowledge diffusion rules.
Many evolution models of knowledge network still have some bias when explaining the evolution
of the real network,which leads to the future direction of evaluating network models. Compared
to the traditional method which directly analyzes the features of the network structure,the method
that uses link prediction to evaluate is more complicated,and applies more advanced mathematical
knowledge,yet it is more convincing in the quantitative comparison level.In the future research,the
evolution model and link prediction should be integrated into one research framework,such as:1)
When the network evolution mechanism is unclear,the main features that affect the network
connection can be found through statistical analysis of the real world network and observation
of link prediction; 2)Based on these features and the previous research findings,the evolution
model of the specific network can be constructed,even its implied rules can be revealed; 3)Use
link prediction to design a quantitative evaluation method,which can evaluate advantages and
disadvantages of the constructed network model and its evolution mechanism.