A comparative study of augmented features and other ensemble approaches for music genre classification

No Thumbnail Available
Shariat, Raad
University of Lethbridge. Faculty of Arts and Science
Journal Title
Journal ISSN
Volume Title
Lethbridge, Alta. : University of Lethbridge, Dept. of Mathematics and Computer Science
Music genre classification is essential in the music streaming industry, with many recommendation systems relying on data mining techniques to accurately classify musical genres. However, classifying music genres is challenging due to the inherent diversity of music, even within a single genre. This diversity can make it difficult for machine learning models to classify music accurately, leading to the development of various techniques to improve the performance of these models. One such technique is ensemble learning, which combines the predictions of multiple models to improve the overall accuracy of the ensemble. In this thesis, we propose a new ensemble method called "Augmented Features," which combines the predictions of multiple models by augmenting the input features with additional derived features. This technique can improve the performance of ensemble models by providing additional information to the models, allowing them to capture the music data's complexity better. To evaluate the performance of our ensemble method, we conducted experiments on various music datasets combined with different feature selection techniques. We compared the results to those obtained using the base classifiers and other ensemble methods, including voting, blending, and stacking. Our results showed that the augmented features method repeatedly outperformed the different techniques, particularly on datasets with high dimensionality and complex relationships between features. It is hoped that this work will significantly contribute to ensemble methods and improve the performance of machine learning models in various applications.
music information retrieval , music genre classification , machine learning , ensemble learning , feature selection technique