Food security and nutritional fulfillment are crucial indicators of regional welfare, particularly in developing regencies like Majalengka. Traditional methods of mapping food vulnerability rely heavily on resource-intensive field surveys, which often lead to delayed interventions. This study proposes a machine learning approach using the Random Forest algorithm to classify the level of food vulnerability across 26 districts, utilizing a time-series secondary dataset (2021-2025) from the Central Bureau of Statistics (BPS). The methodology includes rigorous data preprocessing using the Pandas library to handle missing values, classification modeling using Scikit-learn with 5-Fold Cross Validation, and comparative performance evaluation against Decision Tree and Support Vector Machine (SVM) algorithms. The evaluation results demonstrate that Random Forest achieves an accuracy of 89.23%, outperforming other models in terms of predictive stability and variance reduction. Furthermore, feature importance extraction reveals that agricultural production and population density are the most critical determinants of regional vulnerability. The expected result is an automated Early Warning System capable of mapping nutritional intervention priorities accurately, providing data-driven recommendations for local government policy-making.
Copyrights © 2026