Clustering TV Viewing Behavior for Digital Twin Construction Using Television Viewing History Data

This study presents the construction of the first digital twin utilizing non-identifiable television viewing history data. As the media landscape continues to evolve, understanding viewer behavior has become increasingly crucial. By simulating viewing behaviors based on real-time data, our approach enables the virtual reproduction of viewer preferences and behavior patterns, facilitating optimized advertising, content production, and marketing strategies. We propose a method for classifying user viewing tendencies using large-scale, non-identifiable data and develop a simulator based on these classifications. A detailed analysis of the data led to the extraction of tailored features for television viewing and the development of a highly accurate classification model. The weekday and weekend models achieved F1 scores of approximately 0.95, demonstrating their strong predictive capabilities. This study provides valuable insights into digital twin construction for television viewing and opens new avenues for data-driven media strategies.