计算机科学与探索 ›› 2016, Vol. 10 ›› Issue (2): 201-209.DOI: 10.3778/j.issn.1673-9418.1506047

• 网络与信息安全 • 上一篇    下一篇

面向学术社交网络的多维度团队推荐模型

袁成哲1,曾碧卿2,汤  庸1+,王大豪1,曾惠敏1   

  1. 1. 华南师范大学 计算机学院,广州 510000
    2. 华南师范大学 软件学院,广东 佛山 528225
  • 出版日期:2016-02-01 发布日期:2016-02-03

Multi-Faceted Team Recommendation Model for Academic Social Networks

YUAN Chengzhe1, ZENG Biqing2, TANG Yong1+, WANG Dahao1, ZENG Huimin1   

  1. 1. School of Computer, South China Normal University, Guangzhou 510000, China
    2. School of Software, South China Normal University, Foshan, Guangdong 528225, China
  • Online:2016-02-01 Published:2016-02-03

摘要: 学术社交网络的出现改变了传统的科研方式,对于如何基于学术社交网络为学者进行团队个性化推荐进行了研究,提出了一种多维度潜在团队推荐模型(multi-faceted team recommendation,MFTR)。该模型首先通过投影梯度非负矩阵分解方法提取团队和用户的特征向量,并根据两者的特征向量计算其相似度,然后再融合用户的社交好友关系和热门团队信息来为用户推荐具有相似研究兴趣的潜在团队。最后在真实学术社交网站——学者网的数据上进行实验,结果表明该模型能有效地提高推荐的准确度,并缓解了冷启动问题。

关键词: 学术社交网络, 团队推荐, 非负矩阵分解, 多维度

Abstract: Traditional research methods have been greatly changed by academic social networks (ASNs), this paper explores the area of personalized team recommendation for scholars in ASNs, proposes a novel model named multi-faceted team recommendation (MFTR). MFTR recommends latent scientific research teams with similar research interest for users not only by computing the similarities of users and scientific research teams based on their eigenvector which is abstracted by methods of projected gradient non-negative matrix factorization, but also combining the relationship of friends and active scientific research teams. MFTR is conducted by comprehensive experiments on real world datasets from SCHOLAT that is an academic social site. The results show that the model can efficiently improve the quality of recommendation and abate the problem of cold start.

Key words: academic social networks, team recommendation, non-negative matrix factorization, multi-faceted