This study aims to decipher the dynamic risk evolution mechanism in urban–rural integration development public projects (URID-PPs). It addresses the limitations of static assessments by analyzing how risks propagate across planning, implementation and operational stages to enhance project resilience.
A stage-dependent analytical framework is developed. Large language models (LLMs) are integrated with grounded theory to systematically identify risk factors. To address uncertainty in expert judgment, ELICIT-based linguistic processing is employed, and a hybrid fuzzy BWM and dynamic fuzzy DEMATEL approach is used to quantify factor importance and capture evolving causal relationships across the project lifecycle.
Results reveal distinct evolutionary trajectories: institutional risks dominate the planning phase as root causes, while coordination friction escalates during implementation. In the operational phase, risks shift toward endogenous market mechanisms. Eleven critical risk factors, such as inadequate cross-departmental coordination, are identified, exhibiting stage-specific causal behaviors.
This study advances the research on URID-PPs by moving beyond conventional static risk assessments and introducing a stage-dependent perspective that captures the temporal heterogeneity of risk evolution. By integrating LLM-assisted qualitative analysis and ELICIT information expression, the proposed approach enhances both the systematic identification of risk factors and the rigorous handling of uncertainty in expert judgment. Furthermore, the study reveals how risk causal structures dynamically evolve across different project stages, offering a more nuanced understanding of risk propagation mechanisms and providing an analytical basis for adaptive, stage-specific governance in complex public projects.
