<![CDATA[Stunting mitigation requires precise interventions, yet local health centers frequently face fragmented data. This study develops a preliminary WebGIS-based decision-support prototype for stunting risk prediction to facilitate targeted resource allocation. Utilizing the CRISP-DM methodology, we implemented a direct cross-region model deployment. A Random Forest classifier, trained on a feature-complete perinatal dataset (n = 78) from a source village, was deployed to a target domain integrating 83 toddler records and 10,113 household-level STBM environmental records in Ngawen District. The model achieved an overall accuracy of 81.93% and a weighted F1-score of 0.817. Class-specific F1-scores reached 0.906 (Normal), 0.744 (Mild), 0.757 (Moderate), and 0.818 (Severe). Feature importance analysis identified Birth Weight, Birth Length, and the aggregated village-level STBM score as primary predictors. Furthermore, spatial analysis revealed predicted high-risk clusters in Sarimulyo (5.58%) and Gondang (5.15%), demonstrating an inverse relationship between sanitation coverage and stunting severity. However, these spatial findings are based on model predictions and aggregated indicators rather than confirmed causal relationships. Due to the limited sample size and the prototype's untested status with end-users, broader external validation, clinical verification, and formal usability testing are essential before operational deployment.]]>
Copyrights © 2026
