University Security Situation Assessment and Potential Risk Prediction Based on Spatiotemporal Graph Attention Network

With the expansion of university scale and the complexity of campus environment, campus safety has evolved into a comprehensive system involving multiple dimensions such as personnel flow, facility operation, and emergencies, including the management of laboratories, public facilities and high-density activity areas. The traditional security assessment methods for universities have limitations such as insufficient consideration of spatiotemporal correlations, weak ability to capture hidden risk patterns, and low prediction accuracy, which make it difficult to meet the needs of intelligent and proactive security management. In response to the above issues, this article proposes a university security situation assessment and risk prediction method based on Spatiotemporal Graph Attention Network (ST-GAT). Firstly, establish a multidimensional safety indicator system for universities that covers four core dimensions: personnel safety, facility safety, environmental safety, and emergency management; Secondly, establish a spatiotemporal graph structure to model the spatial correlation between campus functional areas and the temporal dynamics of safety factors; Once again, design the ST-GAT model, integrate graph attention mechanism to enhance the weights of key spatial nodes, use gate controlled cyclic units to capture longterm temporal dependencies, and achieve comprehensive evaluation of security situations and accurate prediction of potential risks. The graph representation can incorporate access-control, camera and wireless sensing nodes when campus security data are available.