期刊引文网络拓扑结构与信息流量的关系研究

Research on Relationship between Structure and Information Flow in Journal Citation Networks

期刊引文网络是典型的信息流网络。期刊在引文网络中的位置不仅反映了期刊当前的重要性和对整个科学系统的贡献,而且对研究人员未来的投稿和引用行为都有重要影响。流量网络的文献中,介数（Betweenness）常被直接或稍加修正后作为真实流量的近似值。比较11个期刊引文网络的实际流量模式和基于介数估计的流量模式后,发现这两种流量模式存在本质差异。本文提出一种根据网络拓扑估计流量的新方法：估计边上流量时假定任意两个顶点间传输的流量与两个顶点的信息需求之积成正比,与顶点之间的网络距离成反比。基于该方法得到的流量网络再现了期刊引文网络真实流量的概率密度分布和流量分布的局部异质性等宏观特征。

Journal citation networks are typical information flow networks.The position of a journal reflects its importance and contribution to the scientific community and exerts some influences on the delivering and citation behaviors of authors in the future.Betweenness is often taken as a direct or slightly modified approximation of real-life flow in studying traffic networks.We test this assumption by comparing the estimating flow pattern with the real-life flow pattern of 11 journal citation networks.We find fundamental difference between those two flow patterns.Then,we propose a new algorithm to estimate information flow by assuming that the traffic flow among any two nodes is in proportion to the product of their information requirement and in inverse proportion to their network distance.The resulting flow networks reproduce most real-life flow patterns of probability density distribution of flow and local heterogeneity of flow distributions.