基于二部配置模型的网络用户分组及重要回复者识别

Internet user grouping and important responder identification based on bipartite configuration model

在以社交媒体为载体的舆论场中,用户的个人诉求和回复行为潜在地反映其兴趣偏好,根据这一关联为用户分组,从而推动群体新闻推荐系统的发展.识别用户的重要回复者以由消息传播机制干预正负面新闻的传播.以天涯论坛为例,定义顶级用户与普通用户,结合二部配置模型建模顶级用户、帖子和普通用户间基于发帖关系以及回复关系的二部图,获取了具有不同话题偏好的用户群,每组内的用户交互密切.得到的重要回复者回复偏重性明显,且易将相应顶级用户的负面情绪放大,因此识别重要回复者并适时施以制约为网络舆情管理提供新视角.

In the public opinion field with social media as the carrier,users’personal appeals and response behavior potentially reflect their interest preferences.Grouping users according to this association helps to develop group news recommendation system.This paper also identifies the users’important responders to intervene in the dissemination of positive and negative news by the message dissemination mechanism.Selecting the case of Tianya forum,the study defines the top users and ordinary users.Bipartite configuration models are used to model bipartite graphs among top users,posts and ordinary users based on the posting relationship and reply relationship.The results show that users in the same group have similar topic preferences,and their interactions are close.Important responders thus obtained are inclined to corresponding top users in terms of the number of replies.The negative emotions of the top users tend to be amplified by their important responders.Thus,the identification of important responders and timely restriction provide a new perspective for online public opinion management.