基于科研网络的潜在合作关系预测研究综述  

A Review on the Prediction of Potential Collaboration Based on Scientific Network

作　　者：方思越 王学昭[1,2] FANG Siyue;WANG Xuezhao(National Science Library,Chinese Academy of Sciences,Beijing 100190,China;School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]中国科学院文献情报中心,北京100190 [2]中国科学院大学经济与管理学院,北京100190

出　　处：《情报工程》2022年第1期35-45,共11页Technology Intelligence Engineering

基　　金：中国科学院战略研究专项:面向国家战略需求的重大科技问题清单研究(GHJ-ZLZX-2021-22-1);国家科技图书文献情报中心资助:全球院士数据建设及支撑高端人才决策场景的情报产品研究(E1510401)。

摘　　要：[目的/意义]合理预测科研领域的潜在合作关系有助于优化资源配置,提升科研产出效率。从科研网络出发的潜在合作预测研究日益增长,需要系统总结。[方法/过程]在CNKI和Web of Science中检索并筛选出基于科研网络的潜在合作关系预测方法的研究,从年发文量、期刊分布对目标文献集进行统计分析。使用内容分析法,梳理出预测潜在合作关系的一般流程,描述步骤中的方法。[结果/结论]潜在合作关系预测一般流程为网络构建、特征提取与表示、合作预测和预测结果评价,其中构建的网络可分为同质网络、异质网络和二分网络,特征提取和表示可分为节点内容特征和网络结构特征,合作预测的方法主要有基于相似性的方法和基于机器学习的方法,预测结果评价的指标为AUC、Precision和Ranking Score;现有方法的局限性启示了未来潜在合作关系预测的发展方向。[Purpose/Significance]The prediction of potential collaboration in scientific research field helps to optimize resource allocation and improve the efficiency of scientific research output.The research on potential collaboration prediction based on scientific network is increasing and needs systematic summary.[Methodology/Process]Studies involving potential collaboration prediction methods based on scientific network were searched and screened in CNKI and Web of Science,and the target literature set is statistically analyzed in terms of annual publication volume and journal distribution.Using content analysis,the general process of predicting potential collaborative relationships is sorted out,and methods are described in each step.[Results/Conclusions]The general process of potential collaborations prediction is network construction,feature extraction and representation,collaboration prediction and results evaluation,where the constructed network can be divided into homogeneous network,heterogeneous network and dichotomous network,feature extraction and representation can be divided into node content features and network structure features,the methods of collaboration prediction are mainly similarity-based methods and machine learning-based methods,and the prediction results evaluation The indicators of the prediction results are AUC,Precision and Ranking Score;the limitations of the existing methods reveal the future development direction of potential collaboration prediction.

关 键 词：科研合作 计量分析 网络科学 

分 类 号：G353.11[文化科学—情报学]