Provincial Segmentation Based on K-Medoids Clustering for Risk Factor Mapping and Targeted Intervention of Stunting in Indonesia

Dina Eka Putri (Dept. of Statistics, Universitas Mataram)
Syifa Salsabila Satya Graha
Baiq Fitria Rahmiati

Journal of Computer Science and Informatics Engineering (J-Cosine)
Vol 10 No 1 (2026): June 2026

Stunting is a multidimensional, chronic nutritional problem shaped by maternal health, caregiving practices, and environmental conditions. This study segments all 38 Indonesian provinces based on ten stunting risk indicators from the 2023 Indonesian Health Survey (SKI) using K-Medoids Partition Around Medoids (PAM) with Median-MAD robust scaling and Manhattan distance. Four optimal clusters were identified (Average Silhouette Width, ASW = 0.425). Sanitation access structurally separates the Papua highlands cluster. Four distinct provincial profiles emerged: (1) good WASH, moderate ANC and breastfeeding gaps; (2) good WASH with high maternal nutritional burden; (3) high anemia despite good service coverage; and (4) severe WASH deficit with low ANC. Cluster-specific policy recommendations are aligned with Indonesia’s existing programmes, i.e. Asta Cita 4, Gerakan Nasional Percepatan Perbaikan Gizi (Gernas PPG), the Thousand Days of Life (HPK) programme, and SDG 2, with targeted enhancements addressing identified gaps.