FOOD AND BEVERAGE PRODUCT SEGMENTATION BASED ON NUTRITION FACTS USING THE DBSCAN METHOD

Abstract

Type 2 diabetes mellitus is increasingly affecting not only teenagers and adults in Indonesia but also children. This serious issue is linked to high-sugar foods, particularly candy and chocolate products consumed by children. The aim of this research is to categorize these products based on their nutritional information, specifically total fat, saturated fat, sugar, and salt (SSF) content per serving, using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method. By doing so, the study seeks to produce simplified product labels that offer clearer nutritional insights compared to conventional nutrition facts labels. Data was collected through purposive sampling from three retail stores. The clustering results, using parameters Eps 0.4 and MinPts 10, revealed two distinct clusters and 133 noise points. Cluster 1 consists of 215 products with low levels of total fat, saturated fat, sugar, and salt, while Cluster 2 includes 27 products that are high in these nutrients. The clustering quality is validated with a Silhouette Coefficient of 0.77 and a Davies-Bouldin Index of 0.345.