Optimizing Music Genre Classification: A Hybrid Approach with ACO and Ensemble Learning
A Sai Kiran Reddy, V Likitha, P Venkata Durga Prasad, V Naga Bharath, S Venkatrama Phani Kumar, Venkata Krishna Kishore Kolli · 2024
The classification of music genres accurately is an ongoing challenge due to the diverse and complex nature of music. Current methods can struggle to identify genres correctly. This study introduces a combination of ensemble learning and Ant Colony Optimization (ACO) to better select features that are key to identifying music genres. Using a diversified collection of songs from Spotify, the research applies detailed data preparation before utilizing a Stacking Classifier framework, which brings together several base learners and an Extra Trees algorithm to improve prediction accuracy. This method achieved a $\mathbf{9 6. 7 5 \%}$ accuracy in classifying music genres, surpassing previous methods and advancing the field. The study also compares this new model with other innovative models, highlighting the effectiveness of using a mix of methods. These improvements have important implications for building better music recommendation systems that are both accurate and user-friendly, paving the way for future developments in automatic music analysis.