Music genre classification is a vital task in the organization and management of large music libraries, providing foundational support for applications such as streaming services, content discovery, and automated DJing. The ability to accurately categorize music into genres enhances user experiences by enabling personalized recommendations and efficient music retrieval. Despite the advancements in machine learning, achieving high accuracy in music genre classification remains challenging due to the complexity of music data and the subjective nature of genre definitions. Traditional models often struggle to capture the nuanced patterns in audio features that are indicative of different genres. This study employs a comprehensive approach to tackle music genre classification by integrating both classical machine learning models and advanced deep learning architectures. We further explore K-means clustering as an exploratory tool to uncover underlying patterns in the dataset. The methodology encompasses data preprocessing, feature engineering, model training, and rigorous validation techniques to ensure robustness and accuracy. The findings reveal that the Dense Neural Network outperforms classical models, achieving an accuracy of 0.650 and a ROC AUC of 0.998. Logistic Regression displayed exceptional discrimination ability with a ROC AUC of 0.999. K-means clustering provided insightful visualizations into the intrinsic clustering of music based on tempo and spectral features, facilitating a deeper understanding of the feature space. The successful application of these models has significant implications for the enhancement of music categorization systems. The integration of deep learning models, in particular, suggests a pathway toward more nuanced and accurate genre classification, which could transform music library management and recommendation systems. These results also underscore the potential for using unsupervised learning techniques to inform feature engineering and model refinement.