Community resilience is a complex, multi-dimensional phenomenon that emerges from nonlinear interactions among diverse features within socio-technical systems. However, most existing assessment methods rely on index-based approaches that inadequately capture the heterogeneity of these systems and their dynamic interdependencies in shaping key resilience components, namely robustness, redundancy, and resourcefulness. To address this gap, this study introduces Resili-Net, an integrated three-module deep learning framework for rating community resilience. The model incorporates twelve measurable features across infrastructure, facility, and social systems, using publicly available data from multiple metropolitan statistical areas in the United States. Resili-Net classifies spatial areas into five distinct resilience levels, and its interpretability allows for identifying the key factors influencing resilience across communities. Scenario analyses simulate how changes in socio-technical conditions affect resilience levels, and a combined resilience–risk analysis reveals communities facing both high hazard risk and low resilience. These findings highlight priority areas for targeted interventions. By modeling resilience as an emergent property of interacting community features, Resili-Net offers a scalable, data-driven approach for resilience assessment that advances current practices through the use of machine learning and heterogeneous urban data.
Yin et al. (Wed,) studied this question.