Comparative Analysis of K-Means and K-Medoids to Determine Study Programs

Abstract

Education is the main foundation for the advancement of civilization. A high level of education in society is directly proportional to that civilization's progress. Higher education is vital in shaping quality human resources and contributing to community and national development. In today's era of information and technology, data processing and analysis are key to understanding the development of study programs in higher education institutions. Clustering techniques identify patterns and relationships in large and complex datasets, which are crucial in determining study programs at educational institutions. This research compares two popular clustering methods, K-Means and K-Medoids, to assess study programs. The data consists of odd semester grades of 87 students in the third year of high school with five variables. The cluster information is based on the minimum academic criteria of 18 study programs representing 7 faculties in Malikussaleh University and is grouped into 5 clusters. The evaluation of clusters is conducted using the Davies-Bouldin Index (DBI). The result of the study indicates that the K-Means algorithm has 5 clusters with cluster members of 31, 5, 13, 26 and 17, and a DBI value of 1,19010. Meanwhile, the K-Medoids algorithm has 5 clusters with cluster members of 33, 15, 17, 17 and 5, and a DBI value of 1,27833. Based on the DBI value, the K-Means algorithm demonstrates better cluster quality than the K-Medoids algorithm.