Dynamics and influences analysis of public concerns in mega construction projects: An integrated topic and sentiment modeling approach

Mega construction projects (MCPs) are often subject to complex and evolving public concerns, with both thematic focus and sentiment attitudes evolving continuously throughout the project lifecycle. This study proposes an integrated topic and sentiment modeling approach to uncover the evolutionary patterns of public concerns and their emotional fluctuations. First, a continuous-time dynamic topic model (cDTM) is developed to extract and track key topics of public concern from unstructured textual data over time. Second, the performance of shallow learning, deep learning, and large language models is systematically compared for sentiment analysis, with the optimal model selected to quantify public sentiment at different phases. Third, a topic-sentiment coupling model is constructed to comprehensively evaluate the critical factors driving public attention and emotional trends, and propose targeted public opinion management strategies. Using a real-world MCP as a case study, the study analyzes 1,163 pieces of textual records and reveals clear phase-specific shifts in public concerns as the project progresses. While public sentiment is generally positive, a notable increase in negative emotions is observed during the construction, expansion, and operational transition phases. The proposed integrated modeling framework provides a scientific basis for public opinion management, stakeholder engagement, and the promotion of sustainable development in MCPs.