Modeling the Causal Relationships between Pre-Construction Delay Risks Using Graph Neural Networks

This study presents a framework combining statistical analysis and deep learning to model pre-construction risks within a risk network, enabling more accurate risk management. The analytic hierarchy process (AHP) was used to model dependencies among 30 pre-construction delay factors, utilizing a risk numerical matrix (RNM) constructed from survey responses of 87 experts. Key network metrics, including in-degree, out-degree, the number of reachable nodes and potential sources, eigenvector, and betweenness centrality, were calculated to analyze risk propagation. These metrics were then used to develop a graph neural network (GNN) to identify the most critical causal relationships between pre-construction delay risks. This study contributes to the field by introducing a framework that integrates qualitative and quantitative approaches to analyze the propagation of pre-construction delay risks based on their interdependencies. The findings emphasize the importance of addressing delays in the pre-construction phase and guide stakeholders in prioritizing and mitigating the most critical risks.