基于频繁传播模式的影响群落发现方法

针对传统研究大多基于影响个体挖掘而忽略了影响群落的发现,本文考虑用户之间的频繁传播模式,提出了一种基于频繁传播模式的影响群落挖掘方法。针对群落内部传播模式的多样化,给出了一种信息传播树扩展方法,通过松弛信息传播树有向特性与图扩展方法,将信息传播树转换为连通无向无环图。结合支持度与影响强度,提出了一种新的频繁子图挖掘算法Tribe-FGM,减小模式增长的规模,提高频繁子图挖掘效率。实验采用新浪微博真实数据,在约90万条博文以及对应约64万左右用户的"地震"话题与约31万条博文以及对应约21万左右用户的"两会"话题的数据集上验证了算法的性能和有效性。

Discovering tribe-leaders based on frequent pattern of propagation

A novel scheme of mining tribe-leaders was proposed based on the frequent pattern of propagation.In this scheme,first,a method to expend the information tree is applied to overcome the problem of multi-pattern propagation,in which the information propagation tree is converted into a connected and undirected acyclic graph.Then,considering the support and influent strength,a new frequent sub-graph mining method called Tribe-FGM is proposed to improve the efficiency of the graph mining by reducing the scale of pattern growth.A real dataset from sina microblog was taken in the experiment.The dataset is about topic of "earthquake",which contains 0.9 million posts and 0.6 million users,and the topic of the "two sessions",which contains about 0.31 million posts and 0.21 users.Experiment results validate the proposed scheme.