事件驱动的在线社交群体演化行为预测

Predicting the Event-driven Evolution Behavior of Online Social Groups

[目的/意义]研究在线社交群体的行为规律,预测群体行为演化趋势,是当前社会计算领域一个重要研究问题。这对于事件传播动向掌控、用户群体异常行为检测、公共事件舆论场管控等有重要意义。[方法/过程]面向在线社交群体,以复杂网络理论为基础,构建融合多维特征的社交群体行为模式分析框架,挖掘网络的社团动态演化模式,对比分析事件驱动下社交群体的行为规律,进行社团演化行为预测。[结果/结论]在合成和真实网络中的大量实验结果表明,社团演化行为预测模型具有较高的精度与较强的鲁棒性;在微博网络中,不同类型事件驱动的群体演化行为特性具有明显差异,其中突发事件驱动的在线社交群体演化行为具有更高的可预测性。

[Purpose/Significance] It is an important research issue to study the behavior laws of online social groups and predict the evolution trend of group behavior in the field of social computing.It helps to control event propagation,detect abnormal behaviors of user groups and manage public opinion fields.[Method/Process] On the basis of complex network theory,this paper proposes an analysis framework of social group behavior pattern by integrating multi-dimensional features to discover the dynamic community evolution of the microblog social network.We compare and analyze the behavior laws of social groups driven by events,and predict the behavior evolution trend of social groups.[Result/Conclusion] The experimental results on a variety of synthetic networks and real social networks confirm that the proposed model of community evolution prediction has high accuracy and strong robustness.For the Weibo social networks,there are obvious differences about the behavior evolution characteristics of online social groups in the different types of events,and the group evolution behaviors driven by emergencies are more predictable.