Music genre classification presents a persistent challenge in music information retrieval because different genres often share acoustic characteristics, while variations in recording quality can further complicate automated classification. This study comparatively evaluates the performance of three machine-learning algorithms—Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT)—for automatic music genre recognition. Audio features were extracted from the selected music samples using Mel-Frequency Cepstral Coefficients (MFCCs) through the Librosa library. The extracted features were subsequently used to train the three classification models, with RF employing an ensemble-based learning approach, SVM identifying an optimal separating hyperplane, and DT classifying samples through a sequence of hierarchical decision rules. Model performance was assessed using accuracy, precision, recall, and F1-score. The results indicated that RF produced the strongest overall performance, achieving 97% accuracy, 95% precision, 94% recall, and 92% F1-score. SVM recorded 95% accuracy, 92% precision, 82% recall, and 73% F1-score, whereas DT achieved 79% accuracy, 78% precision, 74% recall, and 61% F1-score. The comparative results demonstrate that RF outperformed both SVM and DT across all evaluated metrics. Therefore, the study identifies Random Forest as the most effective of the three approaches for the investigated music genre classification task, highlighting its potential for improving automated classification within music information retrieval systems. 

Keywords: Music Genre Classification, Music Information Retrieval, Random Forest, Support Vector Machine, Decision Tree, Mel-Frequency Cepstral Coefficients (MFCCs), Librosa.