Enhancing the reliability of social vulnerability assessment to natural hazards through combined machine learning methods: A case study in Vietnam

Nguyen Thuy Linh, Le Ngoc Hieu,
Enhancing the reliability of social vulnerability assessment to natural hazards through combined machine learning methods: A case study in Vietnam,
Natural Hazards Research,
2025,
,
ISSN 2666-5921,
https://doi.org/10.1016/j.nhres.2025.09.004.
(https://www.sciencedirect.com/science/article/pii/S2666592125000800)
Abstract: Community risk from natural hazards depends on hazard, exposure, and vulnerability; therefore, effective risk reduction and resilience-building demand comprehensive scientific evaluations of vulnerability, including social vulnerability (SoV). While machine learning (ML) has been increasingly adopted in recent SoV studies, its integration into a comprehensive SoV assessment framework remains limited. This research contributes both theoretically and practically by demonstrating the potential of employing ML to enhance the reliability and mitigate the subjectivity of SoV assessments, especially in the indicator-weighting process of the Social Vulnerability Index (SoVI). To achieve this goal, the research compares SoV of six regions in Vietnam that were derived from expert-based methods with those generated using ML methods. Data spanning 2003-2023 were collected from government reports and field surveys across 15 social indicators. Three ML methods (Random Forest, Support Vector Machine, and Decision Tree) were applied alongside the Analytic Hierarchy Process (AHP) to assign weights for SoVI and generate SoVs. The AHP-derived weights were compared against ML-derived weights to assess the correlation. The performance of ML methods was evaluated using precision, recall, and F1-score to determine the most suitable method for each region. An in-depth discussion of the integration of ML methods into SoV assessment was conducted to evaluate the results and suggest a framework that improves the accuracy of future SoV research, whether across diverse geographical areas or in studies involving complex multi-indicator systems or limited research periods. The findings from the case study also support local policymakers in long-term investment planning and resource allocation.
Keywords: Social vulnerability; Social Indicators; Machine Learning; Natural hazards; Vietnam