Evaluation of a TV Programs Recommendation using the EPG and Viewer ’ s Log Data

Evaluation results of a TV program recommendation using EPG (Electronic Program Guide) and viewer’s log data are described. Two experiments using the log data of a Japanese video service provider are conducted: (1) an experiment of prediction of TV programs that each viewer usually watched, and (2) a survey of TV program recommendation using the log data and a questionnaire. High precision values are obtained in the first experiment using keyword-based and celebrity-based recommendation methods for terrestrial and satellite broadcasting programs. Meanwhile, the channel & time-based recommendation method for programs in paid channels of the video service provider obtained the high precision in the second experiment. An overview of experiments and results are described.