Abstract

The Classification of Music genres is an absolutely crucial task in music analysis, with applications ranging from user-tailored music recommendation systems to automated music cata- loguing. This paper presents a novel approach to music genre classification using an ensemble ma- chine learning model which combines the strengths of both Artificial Neural Networks and Convolu- tional Neural Networks. Our objective is to efficiently and accurately classify music samples into their respective genres, while maintaining a focus on enhancing the classification accuracy. The ap- proach involves preprocessing of audio samples to extract their Mel-Frequency Cepstral Coefficients (MFCCs), which are robust feature representations for audio-based data. We compile a comprehen- sive JSON dataset containing MFCC features for each music sample. The ANN is employed for initial data processing, efficiently handling a large dataset, while the CNN excels in capturing spatial pat- terns in audio data, enhancing genre classification accuracy. In this paper, we provide a detailed overview of our methodology, including data preprocessing, model architecture, and training proce- dures. We also conduct experiments using the GTZAN dataset, consisting of 1,000 music samples, to assess the performance of our ensemble model approach. The results produced demonstrate the effectiveness of our strategy in achieving the goal accurate genre classification. This research contrib- utes to the improvement of music genre classification techniques, showcasing the potential of com- bining machine learning models to handle the complexity of audio data. The findings have practical implications for music streaming platforms, content recommendation systems, and music catalogu- ing, thus offering new avenues for enhancing user experiences and automating music analysis.