Mood Classification of Balinese Songs with the K-Means Clustering Method Based on the Audio-Content Feature

Abstract

Bali is a province that has a diversity of arts and can not shunt from songs that come from Bali. Music in Balinese songs has a unique character, both in the variations of the tone that builds up a song and the lyrics contained in a Balinese song. Research on the classification of mood with energy and valence features of a song is often done, especially on western songs. Every music that is thought out has emotional energy that radiates and powerfully connects with human psychology. This research wants to explore whether the features used to classify western songs can also classify Balinese songs, which are rich in the sound of musical instruments according to the tastes of the Balinese themselves. Classification of songs is essential, considering that music is related to specific emotions and moods in humans. In this study, the mood classification of Balinese songs is performed using the Spotify API feature, namely energy and valence. Classification using K-means clustering based on energy and valence features is compared with the song mood data from ten respondents and produces the highest accuracy of 32%.