In the mining of microblogging opinion leaders,features were always separately considered,such as users’attrib-utes,network structures and mutual information.Researchers rarely considered the relationship among these features,as well as the topic features of information in microbloggings.Aimed at above problems,this paper proposed an algorithm,named TopicLeaderRank,for opinion leaders mining in microbloggings based on the user-behavior network.First,the algorithm built a user-behavior network by using users’content and social attributes and mutual information about a specific topic.Then it a-dopted the thought of voting in the PageRank algorithm to find opinion leaders.It also considered the influence from the weights of nodes and edges in the use-behavior network on the voting.Experiments based on three data sets of Sina MicroBlog show that the algorithm is efficient.It gets higher scores on indicators of coverage and coreratio than the sorting algorithm by user’s weights and Microblog-Rank algorithm.It also performs better than the other two algorithms in the artificial evaluation.%微博意见领袖挖掘中通常单独考虑用户属性、网络结构或交互信息等特征，对这些特征之间的关系及微博信息的话题特征考虑较少。针对上述问题，提出了一种基于用户行为网络的微博意见领袖挖掘算法Topic-LeaderRank。该算法利用微博用户的内容属性和社交属性，并结合用户在特定话题中的交互信息构建用户行为网络，然后利用PageRank算法的投票思想，同时考虑网络中节点权重和边权重对投票的影响来挖掘意见领袖。在新浪微博三个话题数据集上的实验结果表明，该算法是有效的，在覆盖度和核心率指标上的值高于用户权重排序和Microblog-Rank算法，在人工评价上的表现也优于这两种算法。
展开▼
