Classic Machine Learning For Mood Classification In Indonesian Music

This research uses datasets of lyrics and audio of Indonesian songs from the 70s and 80s. This research also analyzes the best features of the datasets used. The datasets consists of the chorus of a song which assumes the mood of music lovers. The main music samples selected purposely came from 2 categories, namely pop and ballad/country. The singers of pop songs consist of Chrisye, The Rollies, Fariz RM and Koes Plus. Ballad or Country singers are Ebiet G. Ade and Iwan Fals. Additional music samples selected intentionally were from the Rock and Dangdut/Malay categories. Where the rock singer used is Achmad Albar and the dangdut singer is Rhoma Irama. In the field of psychology, research on mood music aims to find out why humans have emotional responses to music. In the field of Music Information Retrieval (MIR), research on emotion and mood in music aims to create music metadata to make it easier to manage and retrieve music as an entity. The method used in this research is the CRISP-DM method with stages of business understanding, data understanding, data preparation, modeling, evaluation and implementation. This research only reached the evaluation stage. This study uses 1160 data which uses lyrics and audio data and uses 31923 different words. This study uses datasets of Indonesian songs in the 70s and 80s. Where the labels used for mood classification are happy, sad and neutral. The representation of the lyrics used is embedding words and audio consisting of Chroma-gram and Mel-Spectrogram. This research uses several models, namely Naive Bayes, Random Forest, SVM, Catboost and Xgboost which are classic models in machine learning.