Segmentasi Wilayah Provinsi di Indonesia Berdasarkan Indeks Penanganan Stunting Menggunakan PCA dan Partition Clustering

Abstract

Stunting is a public health issue that has long-term impacts on children's cognitive development, productivity, and quality of life. This study aims to cluster provinces in Indonesia based on stunting-related indicators using two partition clustering methods: K-Means and K-Medoids. The analysis involved nine indicators, including immunization coverage, delivery assistance by health professionals, use of modern contraception, exclusive breastfeeding, access to safe drinking water, proper sanitation, food insecurity rate, early childhood education (PAUD) participation, and the percentage of young children from poor families. Dimensionality reduction was performed using Principal Component Analysis (PCA) to improve clustering efficiency and visualization. The clustering performance was evaluated using four internal metrics: Silhouette Coefficient, Davies-Bouldin Index, Calinski-Harabasz Index, and Dunn Index. The results showed that K-Means produced more optimal segmentation than K-Medoids, forming two main clusters. The first cluster includes provinces with strong performance across the indicators, while the second cluster represents provinces with lower performance and higher stunting vulnerability. These findings can serve as a basis for the government to design more targeted stunting interventions tailored to the specific characteristics of each region.