Hybrid Method of Using Feedback from Users to Improve Music

11 Pages Posted: 23 Jan 2023

See all articles by Mashaole Masekwameng

Mashaole Masekwameng

University of the Witwatersrand

Ritesh Ajoodha

University of the Witwatersrand

Date Written: November 23, 2022

Abstract

Automatic music classification relies primarily on traditional aspects of music encapsulated by content-based features. This means cultural aspects of music classification are not incorporated into the classification process. To improve classification, a hybrid method of using feedback from users is proposed to incorporate these cultural aspects. This paper proposes a pipeline of classifying music by genre using content-based features followed by the implementation of Q-learning. The GTZAN dataset is used to model four content-based classifiers, namely: Logistic Regression, Support Vector Machines, Multilayer Perceptron and Random Forests. The highest performing model was the Support Vector Machine (SVM) with and accuracy score of 75.50% based on a 10-fold confusion matrix, in line with other experiments in this field of research. Q-learning is used to further refine the predictions of the SVM by using user input to develop a Q-table with Q-values for each piece of music, associating it with a certain genre based on user feedback. This resulted in an accuracy score of 75.50%, like that of the SVM classification, indicating that Q-learning accuracy cannot overcome the initial classification introduced by the SVM predicted values.

Keywords: Music genre classification, Q-learning, recommendation, Content-based Features, Machine Learning

Suggested Citation

Masekwameng, Mashaole and Ajoodha, Ritesh, Hybrid Method of Using Feedback from Users to Improve Music (November 23, 2022). Available at SSRN: https://ssrn.com/abstract=4332719 or http://dx.doi.org/10.2139/ssrn.4332719

Mashaole Masekwameng

University of the Witwatersrand

Ritesh Ajoodha (Contact Author)

University of the Witwatersrand ( email )

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
29
Abstract Views
185
PlumX Metrics