Adaptive bayesian optimization for sustainable regional resilience: a multiscale framework
开发了一个集成贝叶斯优化框架，结合恢复风险建模与风险感知政策设计，用于评估和增强区域在不利条件下的恢复能力，并应用于欧洲区域数据识别结构性脆弱性。
Abstract This paper develops an integrated Bayesian-optimisation framework to assess and strengthen regional recovery capacity under adverse conditions, combining recovery-risk modelling with risk-aware policy design. To address the joint challenges of sustainability, uncertainty, and performance reliability in complex systems, the study draws together several operations research approaches, including Bayesian variable selection, spatial econometrics, and multi-objective optimisation. The proposed framework combines a dynamic Bayesian shrinkage model to identify the main drivers of recovery across time and space, a discrete-time hazard component to estimate annual and cumulative recovery probabilities, and a policy module that incorporates tail risk and probabilistic constraints through conditional Value-at-Risk and chance-constrained programming. Applied to a large panel of European regions, the framework identifies structural vulnerabilities and supports the design of adaptive interventions under both structural and distributional uncertainty. The results show that recovery reliability is geographically differentiated and sensitive to the recovery horizon, while policy allocations that account for broader structural disadvantage and adverse outcomes yield more balanced and robust intervention profiles.