From Metadata to Meaning: A Hybrid Clustering and Interpretable Rating Analysis of the Netflix Library
Journal of Computing and Social Informatics · 26.03.2026 · 5 (2) · 1-13
DOI: 10.33736/jcsi.11660.2026 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Computing and Social Informatics, online erschienen am 26.03.2026. Verfasst von Siew Mooi Lim und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).
Abstract (Originalsprache)
The rapid growth of streaming platforms has created vast, heterogeneous content libraries, posing challenges for effective content organisation and understanding of audience preferences. This study aims to uncover multi-faceted patterns in Netflix's content structure, categorisation, and audience reception by employing a hybrid analytical framework. Three distinct clustering methodologies—Latent Dirichlet Allocation (LDA), spectral clustering, and K-Prototypes—were applied alongside IMDb rating analysis, genre inference with TF-IDF, and advanced semantic clustering enhanced by interpretable XGBoost and SHAP values. LDA topic modelling identified three distinct thematic areas in content descriptions. Spectral clustering, using Nearest Neighbours affinity and unnormalized Laplacian regularisation, distinguished three clusters based on geographical origin and content maturity. K-Prototypes clustering identified five segments characterised by distinct format, duration, and regional patterns. Furthermore, IMDb rating analysis provided external validation of content quality, genre inference with TF-IDF elucidated textual markers for genres, and semantic clustering revealed high-value genres and low-value tropes. These findings demonstrate that Netflix maintains a carefully balanced portfolio of content spanning different formats, regions, and themes, catering to diverse audience preferences. The complementary nature of the clustering approaches provides a multi-dimensional understanding of streaming content libraries, offering actionable insights for content acquisition strategies, recommendation systems, and data-driven content personalisation on streaming platforms.
Zitieren
APA
Lim, S. M., Chok, X. K., & Choo, Q. X. (2026). From Metadata to Meaning: A Hybrid Clustering and Interpretable Rating Analysis of the Netflix Library. Journal of Computing and Social Informatics, 5(2), 1-13. https://doi.org/10.33736/jcsi.11660.2026
Vancouver
Lim SM, Chok XK, Choo QX. From Metadata to Meaning: A Hybrid Clustering and Interpretable Rating Analysis of the Netflix Library. Journal of Computing and Social Informatics. 2026;5(2):1-13. doi:10.33736/jcsi.11660.2026
BibTeX
@article{Lim2026,
author = {Lim, Siew Mooi and Chok, Xue Kang and Choo, Qi Xiang},
title = {From Metadata to Meaning: A Hybrid Clustering and Interpretable Rating Analysis of the Netflix Library},
journal = {Journal of Computing and Social Informatics},
year = {2026},
volume = {5},
number = {2},
pages = {1--13},
doi = {10.33736/jcsi.11660.2026}
}
EUVIMED