Klasterisasi Lagu K-pop Berdasarkan Suasana Hati Menggunakan Algoritma K-means dan Fitur Audio
ARINA SALMA SALSABILA, Dr. Yohanes Suyanto, M.I.Kom.
2025 | Skripsi | ELEKTRONIKA DAN INSTRUMENTASI
Spotify as a music streaming platform provides various audio features that can be used to understand song characteristics in more depth. The problem in this study arises from the diversity of user emotional preferences and song listening trends, which causes complexity in understanding and grouping music based on mood.
The purpose of this study is to produce a more focused grouping of K-pop songs based on mood, so that users can more easily find songs according to their mood. The method used is the K-means clustering algorithm based on audio valence and energy features. Data of 7.726 K-pop songs obtained from Spotify were grouped into dimensions of four emotion classes in the Russell circumplex model.
The evaluation results showed that the cluster model obtained a Silhouette score of 0,40, a Calinski-Harabasz Index of 7.761, and a Davies-Bouldin Index (DBI) of 0,85.
Kata Kunci : K-means clustering, K-pop, mood, Spotify, audio features