A Machine Learning Approach to Analyzing Main Topics and Sentiments in YouTube's Portrayal of HIV/AIDS in China

Social media creates a hostile environment for HIV health communication, fostering anonymous and fast digital knowledge sharing. However, research on users' digital health comminution on China's HIV narrative on global social media platforms, such as YouTube, remains limited. To advance the communication of health information globally, this study analyzes 255 videos and 3,184 comments (2006-2024) to explore the discourse surrounding China's HIV/AIDS prevention measures on YouTube and provide support for optimizing public health communication strategies. The research employs web scraping techniques to extract metadata and applies methods such as word cloud analysis, keyword analysis, semantic analysis, sentiment analysis, and Latent Dirichlet Allocation (LDA) topic modeling. The results indicate that the main topics of public discussion focus on international cooperation, traditional medicine, and policy transparency. Semantic analysis reveals the impact of cultural factors on public perception. Sentiment analysis shows that positive sentiment is concentrated on prevention effectiveness and international cooperation, while negative sentiment focuses on policy transparency and insufficient information disclosure. The LDA analysis identifies five core themes: international cooperation, cultural factors, policy transparency, the role of traditional medicine, and the effectiveness of public health policies. The findings suggest that public perceptions of China's prevention measures are complex, highlighting the need to optimize communication strategies and improve policy transparency to enhance international recognition and support.