一种基于话题演化的意见领袖发现方法

微博中的意见领袖不仅在社交网络的信息传播中发挥着举足轻重的作用,而且在网络舆情演化中也表现出显著的意见代表性。针对已有的意见领袖挖掘方法仅从复杂网络或者基本图模型来建模发现意见领袖,忽略了意见领袖在具体的话题演化中的意见代表性的问题,提出了基于话题演化的意见领袖发现的方法。该方法首先根据用户之间的交互构建图模型,然后利用寻找中心节点的图论算法挖掘潜在意见领袖,再利用话题演化模型判断潜在意见领袖的演化中心度,最后发现在整体舆情上的具有意见代表性的真实意见领袖。在新浪微博的话题数据集上的试验结果表明,该算法较仅考虑网络模型的意见领袖发现方法更优。

A method of opinion leaders discovering based on the topical evolution

Existing studies about opinion leaders mining only adopted complex network methods or built graph model from networks, which ignored the specific role that opinion leaders play in the evolution of public opinion. To solve this problem, an opinion leaders discovering method based on the topic all evolution was presented. First, this method found latent opinion leaders by building a graph model based on the interaction between users and using the graph theory algorithm of looking for central node. Second, the evolution model of public opinion was used to judge the opinion representation of these opinion leaders. Finally, the true opinion leaders were found out, which had real lead resistance in the overall evaluation of public opinion. Experiment based on Sina MicroBlog datasets showed that this method performed better than those methods only consider the network model.