Abstract: With the rise and popularization of the Internet, social media platforms such as Weibo have become more and more influential in human society. Government microblog is an official account opened by government departments on the microblog platform. It is an important part of the social media platform and an important channel for official communication and guidance of public opinion. It is usually used to publicize government information, respond to public opinion, and promote government achievements. The influence of government micro-blog is the most important measurement index for government departments to carry out policy propaganda and smooth communication between government and people, and selecting the appropriate method to evaluate the influence is the key work for the government to carry out the new media management of government affairs.
The existing work of microblog influence evaluation generally has three modes, based on the basic attributes of users, interactive behavior and social network structure. However, these algorithms usually have some problems such as incomplete index selection, static analysis mode, and single social network construction mode. Therefore, according to the reality of government microblog in social network platform, this paper carries out the influence evaluation of government microblog based on interaction behavior analysis and social network theory. Firstly, we use the Selenium module for the web reptile to obtain the basic information and interaction behavior information of 4492 government microblogs, and the Chinese government microblog data set is constructed. Then, by analyzing the main interactive behavior patterns of microblog (including the interaction between government microblogs, the interaction between government microblogs and ordinary users, and the user recommendation mechanism), the multi-dimensional social network system of government microblogs is constructed according to the mention, forwarding behavior and user recommendation relationship. Secondly, aiming at the defects of the traditional PageRank algorithm, this paper improves an influence evaluation algorithm GM-Rank which is more suitable to government microblog. GM-Rank takes the normalized number of fans and the number of microblogs as the initial PR value, and carries out unequal PR value propagation according to the correlation tightness of microblogs, and iteratively calculates the influence of government microblogs. Finally, the government microblog is ranked and the social network form and development situation of the domestic mainstream government new media are analyzed.
The experimental results show that, compared with the existing microblog influence evaluation algorithm, the GM-Rank method is practical and available, which can intuitively and effectively reflect the influence and development status of government microblog, and also provide a reliable reference for government departments to carry out the evaluation and management of government affairs new media.
Based on the experimental results, this paper analyzes the development status and characteristics of Chinese government microblog, and puts forward three policy suggestions: paying attention to the construction of internal matrix, improving the multi-level system of departments, and strengthening the cooperation and exchange between microblogs. In the next research, we will consider realizing the automation of the whole process through the encapsulation program, in order to serve the construction and development of government microblog more conveniently.

Key words: network public opinion, government microblog, PageRank, social networks, GM-Rank

摘要： 政务微博的影响力是政府部门开展政策宣传,政民沟通情况的重要衡量指标,针对政务新媒体影响力的评估工作也成为政府部门关注的重点。本文基于社会网络理论与交互行为分析开展政务微博影响力评估工作。采用新浪微博的真实数据集,分析政府部门微博的交互行为模式,综合考虑节点的初始影响力、交互行为以及交互强弱程度,根据提及、转发行为以及用户推荐关系构建多维社会网络体系,并在此基础上改进了一种动态影响力评估法GM-Rank,并计算政务微博影响力。结果表明,相较与现有微博影响力评估方法,GM-Rank法更契合实际,可以更直观有效地反映政务微博影响力和发展近况,为政府部门开展政务新媒体的评估与管理工作提供参考。

关键词: 网络舆情, 政务微博, PageRank, 社会网络分析, GM-Rank