Block Clustering Analysis of Users and Content in Digital Radio Listening

In recent years, the mass media industry has increasingly shifted toward a digital-centric model, a trend that isexpected to continue. This study explores the application of machine learning in the digital transformation of massmedia, with a particular focus on modeling and analyzing user behavior in digital radio platforms. Our objectiveis to leverage these insights for advanced business strategies and to address key challenges such as the cold-startproblem. Specifically, we applied clustering techniques to user behavior data. Rather than independently clusteringusers and content, we emphasized the importance of co-occurrence relationships, clustering them jointly. To thisend, we employed block clustering to simultaneously partition users and content into subsets exhibiting distinctinteraction patterns. Based on these clusters, we assigned attribute information to both users and content to enhanceinterpretability. Our analysis revealed that user listening behavior is influenced not only by user attributes but also bythe devices used. For content, factors such as the intended broadcast region and genre significantly impacted listeningpatterns. Moreover, we uncovered distinct behavioral relationships within clusters that jointly include both users andcontent.