Resilience assessment of emergency response capacity in university communities: An integrated approach combining the entropy weight method and BP neural network—A case study of universities in City B

Abstract To implement the holistic national security outlook and the 14th Five-Year Plan’s strategic deployment on enhancing emergency management and intrinsic safety amid urbanization challenges, this study focuses on university communities—entities with dual educational and urban governance functions—to construct a scientific, operational, and scalable resilience evaluation system for their emergency response capabilities. The aim is to identify structural weaknesses and vulnerable links, thereby improving campus safety and stability. Based on resilient city and emergency management theories, we employed literature research, Delphi expert consultation, and empirical analysis to develop a multilevel indicator system comprising three primary dimensions (resistance, adaptability, resilience), nine secondary indicators, and 35 tertiary indicators. The entropy weight method ensured objective weighting, and a BP neural network evaluation model was constructed and trained on MATLAB, optimized through cross-validation and error analysis to achieve high fitting and strong generalization. Empirical testing on five representative universities in City B demonstrated the model’s significant discriminative power in identifying critical dimensions such as dynamic risk screening mechanisms, familiarity with evacuation routes, timeliness of emergency financial aid, and effectiveness of postdisaster psychological interventions. The findings provide systematic methodological support for university emergency capacity building and can be extended to other functional communities (e.g., research institutes, large enterprise parks), contributing to grassroots emergency governance and consolidating the social foundation of national security.