Perbandingan K-Means dan Hierarchical Clustering dalam Pengelompokan Daerah Beresiko Stunting

Abstract

– Currently, Indonesia is one of the countries with a fairly high stunting rate in the world, where the prevalence of stunting is still in the range of 21.6%, while the minimum standard for stunting prevalence set by WHO is 20%. Stunting is a condition of failure to thrive that occurs early in life, usually in children aged 0-5 years. To overcome this problem, the government and related parties have carried out various efforts and intervention programs, one of which is determining priority areas for handling stunting by clustering. In this research, we will cluster stunting areas based on provinces in Indonesia by referring to several parameters, namely the percentage of immunization, proportion of stunting, coverage of exclusive breastfeeding, coverage of vitamins and blood supplement tablets, as well as access to proper sanitation and drinking water. This research will compare clusters formed using Hierarchical Clustering and K Means. The results of the comparison between the K-Means and Hierarchical Clustering methods show that K-Means produces better cluster grouping in terms of the Silhouette Coefficient value of 0.48 and the Calinski-Harabasz index of 10.49 with the number of clusters formed being 2 clusters. In the Hierarchical Clustering Algorithm, the resulting Silhouette Coefficient value is 0.47 and the Calinski-Harabasz index is 9.54. The greater the value of the Silhouette Coefficient and Calinski-Harabasz index, the better the cluster that is formed.