Clustering Sugar Content in Children's Snacks for Diabetes Prevention Using Unsupervised Learning

Diabetes is a chronic health problem with increasing prevalence, especially among children, due to the consumption of sugary foods/beverages. This study aims to cluster children's snack products based on sugar content using unsupervised learning by combining Hierarchical Clustering and K-Means algorithms optimized using Silhouette Score. This combined approach utilizes Hierarchical Clustering to determine the optimal value (????) of K-Means, ensuring the efficiency and accuracy of data clustering. A total of 157 sample data were effectively clustered with K-means. The test results with Silhouette Score yielded the highest value of 0.380 for 2 clusters, while 3 clusters scored 0.350 and 0.277 for 4 clusters. For this reason, 2 clusters better represent the homogeneity of the data in the cluster, although it has not reached the ideal condition. Further analysis showed that high sugar and calorie content in sugary drinks, including milk, could increase blood glucose levels. These findings can be the basis for the development of consumer-friendly nutrition labels. However, support is needed from the government to create a labelling policy to ensure the sustainability of implementation in the community as an educational effort to prevent the risk of diabetes in children.