基于社区结构的科研合作关系分析与预测

Collaboration analysis and prediction in scientific research network based on community structure

科研合作关系分析和预测针对网络的结构信息预测未来哪些学者间会产生合作关系进行研究,对于理解网络信息传播和动态变化具有重要的意义。在主流的基于拓扑属性的关系预测算法基础上提出了一种基于社区结构信息的合作关系预测模型。首先分析社区发现算法下科研网络的链接分布规律及给出模型构建的理论依据,然后构造引入社区拓扑结构信息的改进算法,最后采用不同社区发现算法进行实验。该方法在实验效果和性能上要优于一些经典的算法,说明该算法能够有效地引入社区结构信息对真实的科研合作网络关系预测问题建模,并为科研合作关系分析预测这一问题提供一种新的思路。

Collaboration relation analysis and prediction aims to predict collaboration between scholars according to the network＇s structural information. It is of great significance to information diffusion and dynamic changes. This paper proposed a new algorithm based on the community structure to improve the classic prediction metric which incorporating the topological properties. Firstly,it gave links concurrent distribution under community detection algorithm in collaboration network and theoretical basis for model construction,and then proposed an improved predicting model based on community topology information. Finally,it used different community discovery algorithms for extensive experiment. The experimental results show that the effect and performance of this method is better than some classical algorithms. Moreover,such algorithm can effectively introduce community structure information on the collaboration prediction in real scientific research network and provide a new idea for analysis and forecast of such prediction problem.