Implementasi Metode K-Medoids Clustering untuk Mengelompokkan Kecenderungan Menonton Drama Korea

Abstract

The phenomenon of watching Korean dramas is increasingly growing among female university students and has the potential to create differences in viewing behavior patterns with varying levels of intensity. This study aims to determine the optimal number of clusters using the Elbow method and to identify respondent grouping patterns based on similarities in Korean drama viewing behavior using the K-Medoids method. In addition, this study evaluates the quality of the formed clusters using an internal validation method, namely the Silhouette Coefficient. The data used are primary data obtained through the distribution of questionnaires to female students at Universitas Negeri Gorontalo, incorporating aspects that represent the intensity and tendencies of Korean drama viewing behavior. The analysis begins with determining the number of clusters using the Elbow method based on changes in the Sum of Squared Errors (SSE). The results show that 4 clusters represent the most representative number of clusters visually. Furthermore, the K-Medoids method is applied to group respondents into 4 clusters based on similarities in their Korean drama viewing behavior. However, the evaluation results using the Silhouette Coefficient indicate that the quality of the formed clusters tends to be low. This is reflected in the variation of silhouette coefficient values, where only one cluster demonstrates good clustering quality, while the others exhibit weaker structures with unclear separation between clusters. This condition indicates the presence of data overlap among clusters, resulting in less distinct cluster boundaries