Perbandingan Kinerja K-Means dan K-Medoids dalam Klasifikasi Siswa Berprestasi di SMP Muhammadiyah 60 Medan

Farida Nafisa Siagian
Halim Maulana

Inventa: Journal of Science, Technology, and Innovation
Vol 1 No 3 (2026): : April: Inventa: Journal of Science, Technology, and Innovation

This study investigates the comparative performance of the K-Means and K-Medoids clustering algorithms in classifying student achievement at SMP Muhammadiyah 60 Medan. The research applies a quantitative data mining approach using academic and non-academic variables, including Mathematics scores, Science scores, Bahasa Indonesia scores, attendance records, and extracurricular participation. Data preprocessing was conducted through cleaning, normalization using Min-Max Scaling, and feature selection to ensure data consistency and analytical reliability. The findings indicate that both algorithms successfully classified students into meaningful performance groups with consistent clustering structures. K-Means demonstrated superior computational efficiency and lower SSE values, making it suitable for homogeneous datasets. In contrast, K-Medoids exhibited greater robustness against outliers and produced more stable cluster distributions. The study concludes that K-Medoids provides more representative clustering results for educational datasets characterized by heterogeneous performance patterns.