Clustering Berbasis K-Means dan Pca untuk Segmentasi Spasial Universitas Global Menggunakan Metrik The

Kamalia, Antika Zahrotul
Herlianto, Hemdani Rahendra

Jurnal Teknologi Informasi dan Ilmu Komputer
Vol 13 No 3: Juni 2026

Penelitian ini melakukan segmentasi universitas dunia berdasarkan metrik Times Higher Education World University Rankings (THE WUR) menggunakan K-Means clustering. Dataset hasil web scraping terdiri dari 2.672 entri, dan setelah penanganan ketidaklengkapan data diperoleh 1.907 universitas untuk analisis. Segmentasi dilakukan menggunakan indikator Research, Citations, Industry Income, International Outlook, serta rank, dengan seluruh fitur numerik distandardisasi menggunakan z-score (StandardScaler). Jumlah cluster optimal ditentukan melalui Elbow method dan Silhouette score, sementara kualitas segmentasi divalidasi menggunakan Davies–Bouldin Index(DBI) dan Calinski–Harabasz Index(CHI). Untuk interpretasi, digunakan Principal Component Analysis (PCA) sebagai reduksi dimensi visualisasi dan pemetaan spasial distribusi cluster. Hasil menunjukkan terbentuknya empat cluster dengan profil kinerja yang berbeda, mulai dari universitas riset kelas dunia hingga institusi berkembang yang berorientasi lokal/regional, serta variasi komposisi cluster antar benua yang mendukung kebutuhan peer-based benchmarking dan strategi peningkatan mutu berbasis data.   Abstract This study segmented Global Universities based on the Times Higher Education World University Rankings (THE WUR) metrics using K-Means clustering. The web scraping dataset consisted of 2,672 entries, and after handling incomplete data, 1,907 Universities were obtained for analysis. Segmentation was performed using indicators such as Research, Citations, Industry Income, International Outlook, and rank, with all numeric features standardized using a z-score (StandardScaler). The optimal number of clusters was determined using the Elbow and Silhouette methods, while segmentation quality was validated using the Davies–Bouldin and Calinski–Harabasz Indices. Principal Component Analysis (PCA) was used for interpretation to visualize dimensionality reduction and map spatial distribution clusters. The results show the formation of four clusters with different performance profiles, ranging from World-Class Research Universities to developing institutions with local/regional orientations, and variations in cluster composition across continents, supporting the need for peer-based benchmarking and data-driven quality improvement strategies.