K-MEANS-BASED ACOUSTIC SEGMENTATION FOR MITIGATING POPULARITY BIAS IN MUSIC RECOMMENDATIONS
Abstract
In this research, we propose a Contextual Recommendation System based on acoustic clustering. The model uses the K-Means algorithm (k = 6) to segment a vast Spotify music catalog into statistically distinct sound profiles. Statistical validation by One-Way ANOVA confirmed the distinctness of all six clusters (p < 0.001), ensuring the robustness of the segmentation and revealing the Long Tail structure. The system maps these clusters to usage contexts (e.g., "Focus" → Instrumental Cluster) to mitigate the inherent popularity bias in the data. The approach has demonstrated the potential to increase the discoverability value of recommendations, promoting diversity and addressing the shortcomings of systems based solely on popularity.
References
[1] ADOMAVICIUS, G.; TUZHILIN, A. 2011. Context-aware recommender systems. In Recommender Systems Handbook. Springer, New York, NY, USA.
[2] BURKE, R. 2002. Hybrid recommender systems: Survey and experiments. User Modeling and User-Adapted Interaction 12, 4 (2002), 331–370.
[3] CELMA, O. 2010. Music Recommendation and Discovery in the Long Tail. Springer, Berlin, Heidelberg.
[4] CHARU, C. 2016. Recommender Systems: The Textbook. Springer, Cham, Switzerland.
[5] FISHER, R. 1925. Statistical Methods for Research Workers. Oliver and Boyd, Edinburgh, UK.
[6] GUNAWARDANA, A; SHANI, G. 2015. A survey of accuracy evaluation metrics of recommendation tasks. Journal of Machine Learning Research 10 (2015), 2935–2962.
[7] HUANG, Q.; JANSEN, A.; LEE, J., GANTI, R.; LI, J.; ELLIS, D. 2022. MuLan: A Joint Embedding of Music Audio and Natural Language. arXiv preprint arXiv:2208.12415. https://arxiv.org/abs/2208.12415
[8] JAIN, A. 2010. Data clustering: 50 years beyond K-means. Pattern Recognition Letters 31, 8 (2010), 651–666.
[9] JAIN, T. et al. 2023. Machine Learning Based Music Recommendation System. In 2023 7th International Conference on Intelligent Computing and Control Systems (ICICCS). IEEE, Chennai, India, 105–110.
[10] JHA, M., CHAUDHARY, R.; CHAND, K. 2025. Previsão de músicas do Spotify baseada em aprendizado de máquina. Wiley, Hoboken, NJ, USA, Chapter 8, 145–160.
[11] KAMINSKAS, M.; RICCI, F. 2012. Contextual and Social Music Recommendation. In Proceedings of the 2nd International Workshop on Music Recommendation and Discovery. ACM, Dublin, Ireland, 1–6.
[12] KIM, J.; URBANO, J.; LIEM, C.; HANJALIC, A. 2019. One Deep Music Representation to Rule Them All.? A Comparative Analysis of Different Representation Learning Strategies. Neural Computing and Applications 32, 4 (2019), 1067–1093 DOI: https://doi.org/10.1007/s00521-019-04076-1
[13] LOHMANN, S. 2018. Context-aware content recommendation on news websites. Ph. D. Dissertation. Universität Koblenz-Landau.
[14] MACQUEEN. J. 1967. Some Methods for Classification and Analysis of Multivariate Observations. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability 1 (1967), 281–297.
[15] MEHROTRA, R. et al. 2021. Auditing and Mitigating Popularity Bias in Music Recommendations. In Proceedings of the 2021 Conference on Human Information Interaction and Retrieval (CHIIR). ACM, Canberra, ACT, Australia, 215–224.
[16] MOELANTS, D. 2002. Preferred Tempo for Dance Music. In Proceedings of the International Conference on Music Perception and Cognition. Causal Productions, Sydney, Australia, 400–403.
[17] PARISER, E. 2011. The Filter Bubble: What the Internet Is Hiding from You. Penguin Press, New York, NY, USA.
[18] PERAL, O. 2024. Sistemas de recomendação do Spotify na descoberta e consumo musical. Master’s thesis. Instituto Universitário de Lisboa (ISCTE).
[19] RICCI, F. et al. 2015. Recommender Systems: Introduction and Challenges. In Recommender Systems Handbook. Springer, New York, NY, USA, 1–34.
[20] RUSSELL, J. 1980. A circumplex model of affect. Journal of Personality and Social Psychology 39, 6 (1980), 1161–1178.
[21] SCHEDL, M. et al. 2018. Current challenges and visions in music recommender systems research. International Journal of Multimedia Information Retrieval 7, 2 (2018), 95–108.
[22] SHINDE, S. et al. 2023. Clusterização Baseada em Aprendizado de Máquina Usando Recursos de Áudio do Spotify. In 2023 3rd International Conference on Data Intelligence and Cognitive Informatics in IoT (IDICAIoT). IEEE, Bengaluru, India, 1–5.
[23] SPOTIFY AB. 2024. Spotify Web API: Audio Features Object. https://developer.spotify.com/documentation/web-api/reference/get-audio-features Acesso em: 14 dez. 2025.
[24] THORNDIKE, R. 1953. Who Belongs in the Family? Psychometrika 18, 4 (1953), 267–276.
[25] TINTAREV, N.; MASTHOFF, J. 2015. Explaining recommendations: Design and evaluation. In Recommender Systems Handbook. Springer, New York, NY, USA, 353–382.
[26] TZANETAKIS, G.; Cook, P. 2002. Musical genre classification of áudio signals. IEEE Transactions on Speech and Audio Processing 10, 5 (2002), 293–302.
[27] WAZLAWICK, R. 2014. Metodologia de Pesquisa para Ciência da Computação. Elsevier Brasil, Rio de Janeiro, RJ, Brazil.
License
Copyright (c) 2026 RECIMA21 - Revista Científica Multidisciplinar - ISSN 2675-6218
This work is licensed under a Creative Commons Attribution 4.0 International License.
Os direitos autorais dos artigos/resenhas/TCCs publicados pertecem à revista RECIMA21, e seguem o padrão Creative Commons (CC BY 4.0), permitindo a cópia ou reprodução, desde que cite a fonte e respeite os direitos dos autores e contenham menção aos mesmos nos créditos. Toda e qualquer obra publicada na revista, seu conteúdo é de responsabilidade dos autores, cabendo a RECIMA21 apenas ser o veículo de divulgação, seguindo os padrões nacionais e internacionais de publicação.