关键词: 在线社会网络, 信息传播, 话题推荐, 节点影响力, 动态贝叶斯网络

Abstract: Aiming at the topic recommendation problem in online social networks, this paper focuses on how to find a set of driving nodes which can make the information diffusion broadly, and proposes a new recommendation method that can obtain an approximately optimal set of driving nodes. This method includes three steps：finding the candidate set of driving nodes which have the greatest influence with an extended PageRank algorithm; calculating the breadth of topic diffusion for each driving node in candidate set; and calculating the breadth of topic diffusion for a number of joint driving nodes and finding an approximately optimal set of driving nodes. Experimental results show that the achieved approximately optimal driving node set leads to larger breadth of topic diffusion.

Key words: online social network, information propagation, topic recommendation, user influence, dynamic Bayesian network