Research on Key Technologies for Analyzing Public Opinion Propagation in Social Networks

With the rapid development of the internet, technology, and social media platforms, an increasing number of people are choosing to express their opinions on their social media accounts. This paper aims to address public safety-related public opinion events by analyzing user social networks and node features to predict the extent and paths of opinion dissemination. The study employs a multi-head attention model to extract sentiment features from user tweets, categorizes topics into communities, and integrates updated tweet keywords into node features. It incorporates potential influencing factors from the actual propagation network into the weight matrix of the Graph Attention Network (GAT) to more accurately analyze the impact of neighboring nodes on users and predict the coverage and paths of public opinion dissemination. The research indicates that focusing on the latent factors of public opinion dissemination and node features can provide valuable insights for predicting public opinion spread and designing intervention strategies.