Journal of Frontiers of Computer Science and Technology ›› 2017, Vol. 11 ›› Issue (7): 1056-1067.DOI: 10.3778/j.issn.1673-9418.1607006

Previous Articles     Next Articles

Application of Probabilistic Graphical Model for Discovering User Similarity in Social Network

XU Juan1, ZHANG Di2, QIAN Wenhua3+   

  1. 1. Archives·Party and University History Research Office, Yunnan University, Kunming 650091, China
    2. Personnal Department, Yunnan University of Finance and Economics, Kunming 650221, China
    3. School of Information, Yunnan University, Kunming 650504, China
  • Online:2017-07-01 Published:2017-07-07

概率图模型在社交网络用户相似性发现中的应用

徐  娟1,张  迪2,钱文华3+   

  1. 1. 云南大学 档案馆·党史校史研究室,昆明 650091
    2. 云南财经大学 人事处,昆明 650221
    3. 云南大学 信息学院,昆明 650504

Abstract: Discovering user similarity in social network as a basic research of social media data analysis can be used in product recommendation and social network user relationship evolution effectively. To represent complex correlations and their uncertainties among social network users and improve the accuracy of user similarity in mass social network theoretically, this paper proposes an effective method for discovering user similarity in social network combining network topological structure and dependence between users based on Bayesian network, an important probabilistic graphical model. To improve the scalability of the proposed method and solve the storage and computation problem of mass data, this paper proposes Bayesian network distributed storage and parallel reasoning algorithm based on Hadoop platform. The experimental results verify that the proposed method is effective and correct.

Key words: social network, Bayesian network, user similarity, parallel reasoning, Hadoop

摘要: 社交网络中的用户相似性发现作为社交媒体数据分析中的基础研究,可以应用于基于用户的商品推荐以及社交网络中推导用户关系演化过程等。为了有效地描述社交网络用户间复杂的相关性及不确定性,并从理论上提高海量社交网络用户相似性发现的准确度,研究了基于贝叶斯网这一重要的概率图模型,结合网络拓扑结构和用户之间的依赖程度,发现社交网络用户相似性的方法。为了提高算法的可扩展性,解决海量数据带来的存储和计算问题,提出了基于Hadoop平台的贝叶斯网分布式存储以及并行推理方法。最后通过实验结果验证了算法的高效性和正确性。

关键词: 社交网络, 贝叶斯网, 用户相似性, 并行推理, Hadoop