An Intelligent Spatial Decision Support Framework for Stunting Risk Assessment and Policy Prioritization in Aceh, Indonesia

Abstract

Stunting remains a persistent public health challenge in Aceh Province, where unequal access to nutrition and health services continues to affect district-level child development outcomes. This study proposes an intelligent spatial decision-support framework for stunting risk assessment and policy prioritization under data-constrained conditions. Using 69 district-level observations from 23 districts between 2022 and 2024, the framework integrates predictive analytics, spatial prioritization, and scenario-based intervention analysis to support evidence-informed public health planning. Logistic Regression, Random Forest, and XGBoost are evaluated using calibration-aware metrics, including ROC-AUC, PR-AUC, and recall. Ensemble models achieve the strongest predictive performance, with XGBoost reaching an accuracy of 0.93 and ROC-AUC of 0.89. However, Logistic Regression is retained for primary interpretation due to its stability and interpretability under limited-data conditions. The framework provides probabilistic risk estimates to support district-level prioritization rather than causal interpretation, enabling more adaptive intervention planning under localized public-health conditions. Key risk-related factors include sanitation access, child nutrition services, growth monitoring coverage, adolescent iron supplementation, and exclusive breastfeeding. Scenario-based simulations suggest that coordinated improvements in service access and nutrition programs may reduce predicted district-level risk by approximately 1–4 percentage points in high-prevalence areas. Expert evaluation yields an average usability score of 3.84 out of 5.0, indicating practical relevance for public health stakeholders. The study demonstrates the potential of intelligent spatial analytics to support adaptive and human-centered public health governance.