Abstract: With the rapid development of Internet technology and the popularity of social media, Weibo, as one of the largest social media platforms in China, has become an important source of online public opinion. Due to the large number of Weibo users and the diverse and complex states of public opinion, various factors are intertwined, forming an intricate system. In this system, individual users, Weibos, or comments often play a key role in the evolution of public opinion. Therefore, effectively identifying and analyzing these key elements of public opinion is of great practical significance for monitoring and managing online public opinion.
   Based on hypergraphs, this study constructs a Weibo public opinion hypernetwork analysis model, dividing the Weibo public opinion system into three subnets: social, content, and emotional. The social subnet takes Weibos as hyperedges and users who comment on the Weibos as nodes. The content subnet uses different themes as hyperedges and comments from Weibo users as nodes. The emotional subnet uses emotional intensity as nodes and emotional polarity as hyperedges. This multilayer subnet structure better characterizes the inherent structure and complex relationships of the Weibo public opinion system. To identify key public opinion elements, this study designs a series of algorithms based on hypernetwork characteristics. Firstly, the LDA topic extraction model is used to cluster Weibo comments by theme, identifying theme hyperedges in the content subnet. Secondly, SnowNLP sentiment analysis is employed to calculate the sentiment intensity of Weibo comments, constructing sentiment nodes and hyperedges in the emotional subnet. Finally, key public opinion elements in the social, content, and emotional subnets are identified based on hypernetwork characteristic indicators such as node hyperdegree, hyperedge hyperdegree, hyperedge degree, and information dissemination influence.
   The data for this study comes from real Weibo public opinion topics. By collecting Weibo content, comments, and user information under specific public opinion themes, a Weibo public opinion dataset is constructed. Using “Yuan Longping’s funeral” as a keyword, this study captures Weibo data from November 15, 2021, to November 17, 2021, including 519 valid Weibos, 5,696 comments, and 5,386 comment users. During data analysis, various analytical techniques are employed. Initially, statistical methods are used for quantitative analysis of nodes and hyperedges in the social, content, and emotional subnets, revealing the structural characteristics and distribution patterns of each subnet. Subsequently, key public opinion elements are identified by calculating indicators such as node hyperdegree, hyperedge hyperdegree, hyperedge degree, and information dissemination influence. Finally, sentiment analysis is combined to analyze and discuss the sentiment tendencies of key public opinion elements.
   By constructing a Weibo public opinion hypernetwork model based on hypergraphs, this study uncovers the complex relationships between various elements in the Weibo public opinion system. The theoretical results indicate that the model effectively characterizes the inherent structure and dynamic evolution of the Weibo public opinion system. Simultaneously, the public opinion element identification method based on hypernetwork characteristics proposed in this study demonstrates high accuracy and reliability, providing new ideas and methods for complex network analysis. In empirical research, this study applies the model to real Weibo public opinion topics and identifies six key elements of public opinion: active characters, communicators, hot Weibos, potentially popular Weibos, hot topics, and central themes. The empirical results show that these key elements of public opinion play an important role in the evolution of public opinion. Further analysis reveals that active characters and communicators have high influence and dissemination power in public opinion dissemination; hot Weibos and potentially popular Weibos have more attention and discussions; hot topics and central themes reflect the core content and public opinion trends.
   The application example of this study demonstrates how to apply the Weibo public opinion hypernetwork model based on hypergraphs to actual public opinion monitoring and management. Taking a popular event as an example, this study collects relevant Weibo content, comments, and user information to construct a Weibo public opinion dataset. Then, the constructed model is used to analyze and process the dataset, identify key elements of public opinion, and analyze their characteristics and emotional tendencies. Finally, based on the analysis results, corresponding suggestions for public opinion monitoring and management are proposed. This application example proves the effectiveness and practicality of the research method.

Key words: hypergraph, hypernetwork, Weibo public opinion, key opinion elements, data analysis

摘要： 随着微博用户数日益增多,微博已然成为了网络舆情的重要产生地,同时微博舆情状态多样,涉及因素复杂,多种不同因素构成了一个复杂的系统,系统中存在个别用户、微博或评论等舆情要素对舆情演变起关键作用。因此引入超网络对微博舆情的关键舆情要素进行识别,对网络舆情的分析和监控有积极意义。本研究以基于超图的超网络为基础,构建了微博舆情超网络分析模型,应用LDA,SnowNLP,Python仿真分析等方法,识别微博舆情中的关键舆情要素,并对关键舆情要素的特征和情感进行分析与讨论。最后,应用在真实的舆情主题中,识别出六类关键舆情要素,分别是活跃人物、传播人物、热点微博、潜在热门微博、热点主题、中心主题,并分析各关键舆情要素的情感倾向。研究结果表明,建立的超网络模型,能有效识别特定舆情环境下的关键舆情要素,有利于对网络热门事件进行舆情分析和监控。

关键词: 超图, 超网络, 微博舆情, 关键舆情要素, 数据分析