A Bayesian network model for user´s preference estimation of personalized TV service

In this paper, we propose a statistical method to inference user´s preference on watching TV programs. The inference of preference is one of core modules for the personalized TV service, and we introduce a structure for the service. We designed a signal model for usage history and user preference data, and a statistical model as a Bayesian network, and developed an inference method based on message passing algorithm. With a set of real TV viewers´ watching records at terrestrial TV receiver, we tested our inference model and present the inference results or genre or channel preference given day and time.