Creative Commons source reading · Data quality and algorithm evaluation
AI-driven audience clustering in sport media: a human–computer interaction approach using ‘CoPE-DEC’
Jang YS
Originally published in Front Comput Neurosci · 2026-01-29
From the article
Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.
Abstract
AbstractThis study investigates the characteristics and underlying patterns of sports media audiences from a human–computer interaction (HCI) perspective using artificial intelligence–based deep learning analysis, with the aim of providing foundational data for the sports media industry. To this end, a novel unsupervised clustering framework, the Column-conditioned Prototype-Enhanced Deep Embedded Clustering (CoPE-DEC) technique, was employed to model and analyze multidimensional viewer experience data derived from sports media consumption contexts. The analysis identified three distinct audience clusters with differentiated behavioral, attitudinal, and value-oriented characteristics. The first cluster, labeled “Sports Value Orientation,” was characterized by enhanced concentration during sports viewing, promotion of cooperative…
Results
In this study, the proposed CoPE-DEC model was compared with three alternative embedding-based clustering approaches to evaluate the performance of the final clusters. The baseline embedding models included Truncated SVD, Multiple Correspondence Analysis (MCA), and UMAP. For all models, K-means clustering was applied to the re-sulting embeddings to generate cluster assignments and facilitate comparison. For the SVD embedding, a sparse matrix was constructed using one-hot encoding (Peiliang, 1998). Truncated SVD (e.g., d = 32, random_state = 42) was then fit to the entire dataset to obtain linear low-dimensional embeddings. K-means clustering (K = 3, k-means++, n_init = 20, fixed seed) was applied to this embedding space to assign cluster labels. Evaluation metrics, including Silhouette (Hamming), Davies–Bouldin Index (DBI), and…
Discussion
This study aimed to identify latent viewer typologies in AI-driven sports media environments by applying the Column-conditioned Prototype-Enhanced Deep Embedded Clustering (CoPE-DEC) framework from a human–computer interaction (HCI) perspective. Extending beyond descriptive segmentation, the revised discussion provides an attribution-based interpretation of cluster formation by linking observed audience patterns to underlying HCI mechanisms embedded in contemporary digital sports media ecosystems (Figure 2 and Table 3).
Editorial reading guide
- What data created and tested the model, and was evaluation independent?
- What is the cost of a false positive, false negative or out-of-distribution decision?
