Method to Profiling the Characteristics of Indonesian Dangdut Songs, Using K-Means Clustering and Features Fusion

Abstract

There have been numerous studies that discuss profiling for various subjects, including criminal profiling, consumer profiling, and employee profiling, among others. However, song profiling is a relatively rare and underexplored area. In fact, profiling songs can provide us with new insights. Dangdut, one of the most popular musical genres in Indonesia, is a unique blend of musical rhythms from Arabic, Malay, Indian, and local music, and has the ability to captivate listeners and get them dancing and swaying along. In this study, we utilized feature selection techniques and feature fusion in conjunction with the K-Means clustering method to profile 281 Dangdut songs into two groups of clusters, with the best Silhouette score of 0.646. Additionally, we compared our method with non-Dangdut song data and obtained a Silhouette score of 0.549.