基于HITS算法的微博用户可信度评估

以新浪微博为研究平台,在HITS(hyperlink-induced topic search)算法的基础上,提出融合用户交互行为和博文内容的微博用户可信度评估算法。分别构建基于交互行为和基于博文内容的微博用户有向链接图,图中节点表示用户,有向边体现用户基于交互行为或基于内容的指向关系;依据HITS算法计算两种拓扑结构下微博用户的权威度和中心度;以融合的权威度作为度量评估用户可信度。试验采用从新浪微博采集的数据作为测试集合,通过反复训练法获得可信度阈值,绘制不同可信度算法的用户可信度曲线,验证了算法的可行性和有效性。

Evaluation of microblog users' credibility based on HITS algorithm

Based on Sina-Microblog and HITS(hyperlink-induced topic search)algorithm, a new user's credibility algorithm that merged user interactions and blog contents was putted forward. The new algorithm firstly constructed two directed connection graphs based on user interactions and blog contents respectively, where nodes represented users and arcs embodied the direction relationship between users. Authority and hub of these two connected graphs was computed. The fusion authority was adopted as measurement to evaluate user's credibility. The data collected from Sina-Microblog as test set was used to conduct experiments. Threshold of credibility was obtained by repeated training, and then credibility curves of different algorithms were drawn to verify the feasibility and effectiveness of the new algorithm.