Abstract
Music genre classification presents several challenges, including high-dimensional audio data, overlapping genre characteristics, and subjective labeling. Although machine learning has shown promise in addressing these challenges, many studies either focus on a narrow range of models or lack comparative performance insights. In this study, we perform a comprehensive evaluation of ten machine learning algorithms for music genre classification using the widely adopted GTZAN dataset. The models include Naïve Bayes, Support Vector Machines, K-Nearest Neighbors (KNN), Random Forest, Neural Networks, and Extreme Gradient Boosting (XGBoost), among others. Our results demonstrate that XGBoost achieved the highest accuracy at 90.09%, outperforming Random Forest (81.42%) and KNN (80.58%) by a significant margin. Feature importance analysis highlights that percussive, harmonic, and spectral features contribute most substantially to genre discrimination. Notably, XGBoost’s superior performance suggests its ability to effectively capture non-linear patterns in musical features, offering a strong alternative to traditional classifiers. This work contributes a broad comparative analysis and emphasizes the effectiveness of ensemble-based approaches in music genre classification, providing valuable insights for future research and practical applications in music information retrieval.