Klasifikasi Genre Musik Menggunakan Algoritma Support Vector Machine

Music Genre Classification Using Support Vector Machine Algorithm

Abstract

The development of digital music platforms such as Spotify and YouTube Music has led to a significant increase in the volume of music data available online. This situation presents challenges in organizing and retrieving songs that match user preferences. One viable solution is automatic music genre classification using machine learning techniques. This study aims to develop a music genre classification model using the Support Vector Machine (SVM) algorithm based on text features and artist metadata. The dataset used consists of the attributes artist_name, tags, and playlist genre. The research stages include data cleaning, text feature transformation using the Term Frequency-Inverse Document Frequency (TF-IDF) method, artist name encoding using Label Encoder, train-test data splitting, SVM model training, and evaluation using accuracy, precision, recall, and F1-score metrics. To improve model performance, parameter optimization was carried out using GridSearchCV. The results show that the SVM model is capable of classifying music genres with an accuracy of 67%. The Pop genre achieved the best classification performance, while the Rock and R&B genres exhibited relatively high misclassification rates due to data distribution imbalance and the limitations of the features used. The findings indicate that the SVM algorithm is sufficiently effective for text-based music genre classification; however, the addition of audio features and data balancing techniques is still required to further improve model performance.