Multi-Level Analysis of Resilience in Regions Exposed to Environmental Stresses: A Systematic Review of Modeling Approaches and Machine Learning Applications

This paper presents a systematic review of resilience analysis across micro, meso, and macro levels in regions exposed to environmental stresses. Unlike conventional studies, this review focuses not on numerical computations but on the type of information that different modeling approaches—particularly machine learning (ML)—can provide at each level. Based on a structured search of Scopus databases and following PRISMA-based screening, ۴۷ relevant documents were analyzed. The findings reveal that at the micro level, ML models such as random forest explain household adaptive capacity; at the meso level, clustering and social network analysis identify vulnerable communities; and at the macro level, unsupervised learning enables regional typologies of resilience. A conceptual model is proposed to illustrate how information flows across levels without requiring mathematical integration. A key theoretical gap is the lack of integrated cross-scale models. This review provides a conceptual framework for selecting ML models based on the desired output information.

کلیدواژه ها:

Environmental resilience ، levels of analysis ، systematic review ، machine learning ، agent-based modeling ، adaptive capacity ، cross-scale integration