Electronic Theses and Dissertation

Stunting merupakan permasalahan kesehatan yang masih menjadi perhatian di indonesia, termasuk di pulau sumatra, dengan kondisi faktor-faktor penyebab yang beragam antar kabupaten/kota. perbedaan karakteristik wilayah terkait kesehatan ibu dan anak, pemenuhan gizi, serta akses sanitasi dan layanan kesehatan menyebabkan perlunya dilakukan pengelompokan wilayah untuk memahami pola kesamaan karakteristik faktor penyebab stunting. penelitian ini bertujuan untuk membandingkan kinerja metode clustering density-based spatial clustering of applications with noise (dbscan) dan k-medoids dalam mengelompokkan kabupaten/kota di pulau sumatra berdasarkan faktor-faktor penyebab stunting serta menentukan metode dengan kualitas klaster terbaik. data yang digunakan merupakan data sekunder berupa data profil kesehatan tahun 2022 yang diperoleh dari dinas kesehatan tiap provinsi di pulau sumatra, dengan unit analisis sebanyak 135 kabupaten/kota dan 7 variabel. evaluasi hasil klasterisasi dilakukan menggunakan lima indeks validasi, yaitu silhouette coefficient, dunn index, rasio simpangan baku, calinski-harabasz index (chi), dan davies-bouldin index (dbi). hasil penelitian menunjukkan bahwa metode dbscan merupakan metode terbaik dengan parameter epsilon sebesar 1,0 dan minpts sebesar 3, yang menghasilkan dua klaster dan noise. klaster 1 terdiri atas 114 kabupaten/kota dengan karakteristik capaian indikator di bawah rata-rata, sedangkan klaster 2 terdiri atas 3 kabupaten/kota dengan karakteristik di atas rata-rata. selain itu, terdapat 18 kabupaten/kota yang teridentifikasi sebagai noise karena memiliki karakteristik yang berbeda dari klaster utama. berdasarkan hasil validasi, dbscan memberikan performa yang lebih baik dibandingkan k-medoids, ditunjukkan oleh nilai silhouette coefficient sebesar 0,523, dunn index sebesar 0,298, rasio simpangan baku sebesar 0,405, dan dbi sebesar 0,490. hasil penelitian ini diharapkan menjadi bahan pertimbangan bagi pemerintah dalam merumuskan kebijakan intervensi stunting yang lebih terarah sesuai karakteristik wilayah.

Stunting remains a significant health concern in Indonesia, including on the island of Sumatra, with the underlying causes varying across districts and cities. Differences in regional characteristics related to maternal and child health, nutritional status, and access to sanitation and health services necessitate the clustering of regions to identify patterns of common characteristics among the factors contributing to stunting. This study aims to compare the performance of the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and K-Medoids clustering methods in grouping districts/cities on the island of Sumatra based on factors causing stunting and to determine the method with the best cluster quality. The data used consists of secondary data in the form of 2022 health profile data obtained from the Health Office of each province on the island of Sumatra, with 135 districts/cities as the units of analysis and 7 variables related to factors causing stunting. The clustering results were evaluated using several validation indices, namely the Silhouette Coefficient, Dunn Index, Standard Deviation Ratio, Calinski-Harabasz Index (CHI), and Davies-Bouldin Index (DBI). The results showed that the DBSCAN method was the best method with an epsilon parameter of 1,0 and MinPts of 3, producing two clusters and some noise data. Cluster 1 consists of 114 districts/cities with indicator achievement levels that are relatively below average, while cluster 2 consists of 3 districts/cities with indicator achievment levels above average. Additionally, 18 regencies/cities were identified as outliers because their characteristics differed from those of the main clusters. Based on the validation results, DBSCAN performs better than K-Medoids, as indicated by a Silhouette Coefficient of 0,523, a Dunn Index of 0,298, a Standard Deviation Ratio of 0,405, and a DBI of 0,490. It is hoped that the results of this study can serve as a basis for the government in formulating more targeted stunting intervention policies in accordance with regional characteristics.