Clustering Snack Products Based on Nutrition Facts Using SOM and K-Means for Diabetic Dietary Recommendation

The number of diabetics in Indonesia continues to rise, with Type II Diabetes Mellitus (DM) dominating 90% of cases. One of the main contributors is the excessive consumption of snack products high in Sugar, Salt, and Fat (SSF), which increases health risks, particularly for diabetics. However, the current nutrition facts provided in the product package is not easy to understand. Creating label for the product can make an effective information to assist people on buying decision. This study aims to segment snack products based on their nutritional facts, particularly focusing on their SSF content, to identify products that are potentially high-risk for diabetics. In this study, data on the nutritional facts of snack products were analyzed. Utilizing a hexagonal Self-Organizing Map (SOM) topology with a 5 × 9 grid, the best clustering method identified was k-means. This method yielded two clusters, with a silhouette index of 0.44, a Dunn index of 0.09, and a connectivity index of 11.14. The first cluster comprises 165 products that have low levels of total fat, saturated fat, sugar, and salt. In contrast, the second cluster consists of 46 products with high total fat and saturated fat content, and this cluster is of particular concern due to its elevated levels of these unhealthy fats. The segmentation results can serve as a reference for more intuitive food labeling, potentially improving consumer awareness and aiding in dietary decision-making, particularly for diabetics.