Clustering Indonesian Traditional Foods by Nutritional Profiles using the K-Means Algorithm for Health Policy

Irene Devi Damayanti (Department of Informatics Engineering, Universitas Kristen Indonesia Toraja, 91811, Indonesia)
Samuel Yacobus Padang (Department of Informatics Engineering, Universitas Kristen Indonesia Toraja, 91811, Indonesia)
Isak Tandi (Department of Informatics Engineering, Universitas Kristen Indonesia Toraja, 91811, Indonesia)
Melda Duma' (Department of Agrotechnology, Universitas Kristen Indonesia Toraja, 91811, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 10, No 2 (2026): Zero: Jurnal Sains Matematika dan Terapan

Indonesia has a wide variety of traditional foods; however, systematic mapping of their nutritional composition remains limited. This study aims to cluster Indonesian traditional foods based on nutritional profiles using the K-Means algorithm to support health policy development. The analysis focuses on calories, protein, fat, and carbohydrates. A quantitative approach was applied, including data selection, normalization, and determination of optimal number of clusters using the Elbow Method. The results show that four clusters (k = 4) were obtained. Cluster 3 contains foods high in calories and protein, Cluster 2 is dominated by carbohydrates, Cluster 0 shows moderate nutritional values, and Cluster 1 represents low-energy foods. Clustering quality was evaluated using the Silhouette Coefficient (0.45) and Davies–Bouldin Index (0.94), indicating moderate and acceptable clustering performance. PCA visualization retained 88.77% of total data variance. Clusters inform policy via dietary grouping, guiding interventions; limited to macronutrients, excluding micronutrients and portion variability.