<![CDATA[Food security is a strategic and increasingly complex issue in Indonesia, where agricultural productivity, urbanization, climate change, and regional disparities pose major challenges. Data-driven analysis provides significant opportunities to support the formulation of evidence-based food policies, one of which is through clustering methods. This study aims to compare the performance of the K-Means and K-Medoids algorithms in classifying 34 Indonesian provinces based on food security indicators, namely plant energy availability, animal energy availability, total energy availability, and the ratio of animal energy. The dataset was obtained from the Central Statistics Agency (BPS) for 2023. The research process involved data preprocessing, normalization, application of both algorithms, and evaluation using the Silhouette Coefficient, Davies–Bouldin Index (DBI), and the Elbow Method to determine the optimal number of clusters. The results indicate that K-Means with k=5 provides the best balance between internal validity and policy usefulness, while K-Medoids demonstrates better robustness against outliers and higher interpretability through the use of medoids as cluster centers. The comparative results show that K-Means performs better for clearly separated clusters, whereas K-Medoids yields more stable results for data with high variability. These findings highlight the importance of selecting clustering algorithms according to data characteristics and analytical objectives. This study contributes to the development of data-based approaches for food policy formulation and provides a scientific foundation for designing more targeted and effective food security intervention strategies across Indonesian provinces.]]>
Copyrights © 2025
