Computer Engineering and Applications ›› 2022, Vol. 58 ›› Issue (21): 172-181.DOI: 10.3778/j.issn.1002-8331.2103-0540

• Pattern Recognition and Artificial Intelligence • Previous Articles     Next Articles

Collaborator Recommendation Model Based on Academic Level Clustering

QIN Hongwu, ZHAO Meng, MA Xiuqin, ZHAO Dezhi, YAN Wenying   

  1. College of Computer Science & Engineering, Northwest Normal University, Lanzhou 730070, China
  • Online:2022-11-01 Published:2022-11-01

基于学术水平聚类的科研合作者推荐模型

秦红武,赵猛,马秀琴,赵德志,闫文英   

  1. 西北师范大学 计算机科学与工程学院,兰州 730070

Abstract: In order to solve the problem that the existing collaborator recommendation model does not consider the academic level gap between the target scholar and the recommended scholar, this paper proposes an collaborator recommendation model based on academic level clustering , recommend the most suitable collaborators for the target scholars(fitting collaborator recommendation, FCR). FCR uses the K-means algorithm to cluster the scholars in the cooperative network according to the characteristics of the academic level, and then establishes the cooperative network in the same level category, and uses the Katz indicator in the link prediction algorithm to perform similarity to the nodes in the network, and calculates and extracts the research topics of scholars. Finally, the accessibility of the node in the network, whether the academic level is similar or not, and the similarity of research topics are comprehensively considered, comprehensively calculate their recommendation scores and make Top-N recommendations. Compared with other recommendation models, the proposed model has better performance in accuracy, recall and F1 index, which are 5.3%, 2.5% and 4% higher on average. In addition, the academic level matching between the other models has been improved by 37% on average.

Key words: collaborator recommend, social network, K-means, Katz index

摘要: 针对现有科研合作者推荐模型一般不考虑目标学者与推荐学者间学术水平的差距,导致合作关系难以建立的问题,提出一种基于学术水平聚类的合作者推荐模型,为目标学者推荐最合适合作者(fitting collaborator recommendation,FCR)。该模型先使用K-means算法对合作网络中的学者按照学术水平特征进行聚类,在同一水平簇别中建立合作网络,利用链路预测算法中的Katz指标对网络中的节点进行相似度计算,对学者们的研究主题进行提取,在网络的可达性,学术水平是否相近以及研究主题相似度三个方面进行综合考虑并进行Top-N推荐。实验结果表明,相比于其他模型,提出的基于学术水平聚类的合作者推荐模型相比于其他推荐模型均有着较优的表现,在推荐的准确率、召回率以及F1指数上分别提高了5.3%、2.5%、4%。并且在推荐的合作学者与目标学者的学术水平匹配性方面平均提高了37%。

关键词: 合作者推荐, 社会网络, K-means聚类, Katz指标