社交网络高影响力用户发现算法研究

Detecting Influential Leaders Based on Topic Influence Tree in Social Networks

社交网络中高影响力用户在产品推荐、广告营销、舆情监测、信息阻断等方面有着广泛的应用.有效挖掘高影响力用户与社交网络的拓扑结构、用户发布的消息及围绕消息所产生的交互行为都紧密相关.当前的研究主要围绕社交网络节点、用户话题、转发行为等不同维度进行深入分析.如何有效对社交网络短文本消息进行话题描述、探查用户发布消息与用户交互行为之间的级联机制,已成为社交网络影响力用户发现的关键问题.针对上述关键问题,在提取基于词向量描述的话题子网、hashtag标题子网的基础上,提出了基于话题的社交网络影响力用户发现方法,在不同的数据集上,通过与不同的社交网络影响力用户发现算法进行对比,并对基于不同数据集的影响力用户的可视化分析,实验验证了所提算法的有效性.

Identification of highly influential users in social networks is very important in many fields, such as targeted advertisement, recommending products, and tracking hot topics, etc. Information and behaviors of users are two core facets for ranking leaders＇ influence. Current existing studies have been more widely focused on the effect of analyzing influential leaders in terms of topical users. How we can model users ＊ topics revealing the hidden relationship between tweets and behaviors for measuring influential is very important in real social networks. Based on various features, this paper provides a method to model fine-grained topic and capture effects between topics and behaviors in the form of influence tree to measure the influence of users in social networks. Through extensive experiments comparing with different algorithms, and visualization on TUAW corpus, we demonstrate that model is able to identify influential leaders. Experimental results demonstrate the effectiveness of our algorithm.