Claim Missing Document
Check
Articles

A HYBRID BAYESIAN OPTIMIZATION-GENETIC ALGORITHM APPROACH FOR OPTIMIZING OPTICS PARAMETER IN REGIONAL CLUSTERING BASED ON SOCIOECONOMIC INDICATORS Siskawati Zees; Dewi Rahmawaty Isa; La Ode Nashar
Journal of Mathematics, Computations and Statistics Vol. 9 No. 3 (2026): Volume 09 Issue 03 (September 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/gx8yt508

Abstract

Socioeconomic conditions across regencies and cities in Indonesia are highly heterogeneous, making it difficult to identify regions with similar welfare levels using conventional clustering methods. Density-based clustering, particularly Ordering Points to Identify the Clustering Structure (OPTICS), can accommodate varying cluster densities and the presence of outliers; however, its performance is influenced by parameter selection. Although optimization approaches have been used to address parameter selection, previous studies have not specifically applied a hybrid Bayesian Optimization and Genetic Algorithm (BO-GA) approach to optimize OPTICS parameters for regional clustering based on socioeconomic indicators. This study aims to optimize OPTICS parameters and identify socioeconomic clusters of regencies and cities in Indonesia in 2024. The data comprise seven socioeconomic indicators for 514 regencies and cities obtained from Statistics Indonesia (BPS). Optimization was performed using one-way hybridization, in which Bayesian Optimization (BO) was initially used to explore the parameter space of MinPts, ε, and xi and generate the top-k solutions as the initial population for the Genetic Algorithm (GA). Subsequently, GA refined the candidate parameters through an evolutionary mechanism, using the Silhouette Coefficient (SC) as the objective function. The results show that the SC increased from 0.612 during the BO stage to 0.687 after the GA stage, with optimal parameters of MinPts = 13, ε = 20.861, and xi = 0.023. Applying these parameters produced two clusters containing 19 and 15 regencies or cities, respectively, while the remaining 480 regions were classified as noise because they were not assigned to either cluster. The indicator profiles suggest that Cluster 0 exhibits relatively lower socioeconomic welfare characteristics than Cluster 1. An SC value of 0.687 indicates relatively good cluster cohesion and separation. Thus, the BO-GA combination was effective in optimizing OPTICS parameters, although most regions remained unclustered. Future research is recommended to incorporate additional socioeconomic and spatial indicators to provide a more comprehensive representation of regional characteristics.
Perbandingan Metode Mean Shift Clustering Dan Density Peaks Clustering Dalam Pengelompokan Provinsi Berdasarkan Indikator Kesehatan Lingkungan Di Indonesia Alvira Napu; Fahrezal Zubedi; La Ode Nashar
Science and Education Journal (SICEDU) Vol 5 No 3 (2026): Science and Education Journal 2026
Publisher : LPPM Universitas Pahlawan Tuanku Tambusai

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.31004/sicedu.v5i3.1455

Abstract

Kesehatan lingkungan merupakan salah satu aspek penting dalam meningkatkan kualitas hidup masyarakat. Namun, capaian indikator kesehatan lingkungan masih menunjukkan perbedaan antarprovinsi di Indonesia sehingga diperlukan analisis pengelompokan untuk mengidentiikasi provinsi yang memiliki karakteristik serupa. Penelitian ini bertujuan membandingkan kinerja metode Mean Shift Clustering dan Density Peaks Clustering (DPC) dalam mengelompokkan provinsi berdasarkan indikator kesehatan lingkungan di Indonesia. Data yang digunakan merupakan data sekunder Profil Kesehatan Indonesia Tahun 2024 yang mencakup 38 provinsi dengan sembilan indikator kesehatan lingkungan. Analisis dilakukan menggunakan bahasa pemrograman Python. Penentuan parameter bandwidth pada Mean Shift Clustering menggunakan pendekatan quantile, sedangkan parameter cut‑off distance pada DPC menggunakan pendekatan persentil. Kinerja kedua metode dievaluasi menggunakan Silhouette Coeficient (SC) dan Davies‑Bouldin Index (DBI). Hasil penelitian menunjukkan bahwa Mean Shift Clustering menghasilkan dua klaster, sedangkan DPC menghasilkan dua klaster utama dan satu klaster outlier yang merepresentasikan provinsi dengan karakteristik berbeda dari klaster lainnya. DPC memperoleh nilai SC sebesar 0,7257 dan DBI sebesar 0,3342, sedangkan Mean Shift Clustering memperoleh nilai SC sebesar 0,7010 dan DBI sebesar 0,4440. Hasil tersebut menunjukkan bahwa DPC menghasilkan struktur klaster yang lebih kompak dan memiliki pemisahan antarklaster yang lebih baik, sehingga lebih sesuai digunakan untuk pengelompokan provinsi berdasarkan indikator kesehatan lingkungan di Indonesia.