基于网络拓扑结构与节点属性特征融合的科研合作预测研究

Research on Scientific Collaboration Prediction Based on the Combination of Network Topology and Node Attributes

文章从科研合作网络中的作者节点属性出发,提出两种基于合作机构偏好相似性的IDF和ICCR指标,并将其与目前常用的基于网络拓扑结构相似性的CN、AA、LP和Katz指标进行加权融合,构建了8种潜在科研合作关系的预测算法。以化学领域的作者合作网络为研究对象,对8种预测算法的预测效果进行了实证检验,研究发现基于融合性指标的加权预测算法能够达到较好的预测效果,且ICCR指标的表现略优于IDF指标。

Based on the author node attributes in the scientific collaboration network,this article proposes two indicators named as IDF and ICCR,which can reflected the similarity of authors’preferences in choosing cooperation institutions.By combing the two indicators with four commonly used indicators CN,AA,LP and Katz,which are based on the network topology similarity,eight weighted algorithms for prediction of the potential scientific collaboration are constructed.The author collaboration network in the field of chemistry is chosen to verify the prediction effects of these algorithms.It is found that the weighted prediction algorithms based on the fusion indications can achieve better prediction results,and the performance of ICCR is slightly better than IDF.