近年来, 社交网络已成为用户普遍分享信息和交流互动的媒介, 这使得社交 网络中的舆情信息传播具有速度快、覆盖范围广等特征. 但是, 由于存在较多破坏性强、非理 性的负面舆情信息, 故社交网络舆情信源的发现和控制受到了学术界与相关监管者的广泛关注. 文章针对社交网络中节点间关系强度不确定、舆情信源定位困难等问题提出了一种两阶段的舆情信源发现方法, 以传统社交网络SI模型为基础, 融入用户关系强度进行优化, 在异质网络环境下结合概率加权图和宽度优先搜索树进行建模, 并结合Louvain算法进行算法设计, 最后利用BA无标度网络和真实社交网络用户数据集进行算法比较. 实验结果表明, 文章所提舆情信源发现算法从运行效率和准确率来看都优于现有的信源定位算法.

Social network services have become a more common use case of medium for sharing information and communicating interaction, which makes the public opinion information in it with the characteristics of quick to transmit and spread widely. However, for the reason of the existence of more destructive and non-rational negative public opinion information, resulted in the discovery and control to public opinion source in social network services has been widely concerned by academia and relevant regulators. This paper presents a two-stage discovery method of public opinion source for problems where the uncertain relationship strength between nodes of social network services, also hard to locate of public opinion source, and so on. Optimized for integration of user relationship strength, which based on traditional social network services SI models, and in network anomaly, probability plus weighted graph and breadth-search tree are combined to model, also incorporated algorithm designs that combined with Louvain algorithm, in the end, the algorithm comparison is set by using BA scale-free network and real social network user data. The experiment turns out the viewpoint about the discovery algorithm of public opinion source put forward in this paper is superior to the existing source location algorithms in terms of efficiency and accuracy.