Decoding the Subjective Symphony: Machine Learning for Music Genre Classification Using Audio Features and Taxonomy Reduction

V. Ajitha, J. Nishitha, R. Naga Rishika, Md Sufiyan, K. Sai Umanth, K. Laxmi Narayana

2025

Abstract

This paper addresses the application of machine learning to categorize music genres using audio attributes from a dataset of 114,000 songs representing 125 genres. Addressing the difficulty of categorizing detailed, subjective audio data into several genres, the research aimed to predict both individual and multi-genre classifications while decreasing the goal taxonomy from 114 to 56 genres using hierarchical clustering. Through exploratory data analysis, the dataset’s 15 features (11 numeric, 4 categorical) were pre-processed, non-audio genres removed, and data balanced. Models tested included a neural network, Boost, K-nearest neighbours, and an ensemble approach. Evaluated using top-3 categorical accuracy, a critical metric for recommendation systems in which neural network achieved the highest performance (73.74%), followed by the XG Boost (69.54%) and KNN (70.36%). Results revealed that genres with separate aural properties were labelled more accurately than those with overlapping traits, underscoring the subjective complexity of musical categorization. The study implies that while machine learning shows progress, genre subjectivity remains a basic hurdle. Future directions include refining ensemble techniques, adding lyrics, and exploring multi-label classification to enhance accuracy and nuance in music categorization.

Paper Citation

in Harvard Style

Ajitha V., Nishitha J., Rishika R., Sufiyan M., Umanth K. and Narayana K. (2025). Decoding the Subjective Symphony: Machine Learning for Music Genre Classification Using Audio Features and Taxonomy Reduction. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 762-766. DOI: 10.5220/0013920400004919

in Bibtex Style

@conference{icrdicct`2525,
author={V. Ajitha and J. Nishitha and R. Rishika and Md Sufiyan and K. Umanth and K. Narayana},
title={Decoding the Subjective Symphony: Machine Learning for Music Genre Classification Using Audio Features and Taxonomy Reduction},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={762-766},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013920400004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - Decoding the Subjective Symphony: Machine Learning for Music Genre Classification Using Audio Features and Taxonomy Reduction
SN - 978-989-758-777-1
AU - Ajitha V.
AU - Nishitha J.
AU - Rishika R.
AU - Sufiyan M.
AU - Umanth K.
AU - Narayana K.
PY - 2025
SP - 762
EP - 766
DO - 10.5220/0013920400004919
PB - SciTePress