Content-based music classification using ensemble of classifiers

Abstract

This paper presents an application of Ensemble learning in the field of audio data analytics. We propose a system using Hierarchical ensemble model to classify the genre of a music track based on the contents of the track. The hierarchical ensemble comprised of 7 classifiers trained on different sections of the dataset that can co-relate the output of each other for classifying the data. Using this hierarchical ensemble model, we achieved an accuracy boost of 15% over machine learning models. This hierarchical ensemble has been proven better than an ensemble model with hard voting logic in term of accuracy. This work describes the comparison of basic models with hierarchical model and its characteristics.