Mining Association Rules from TV Watching Log for TV Program Recommendation

Abstract

This paper proposes a method for extracting association rules from users’ TV watching logs, aiming at TV program recommendation. In Japan, a TV is usually located in a living room, where family members communicate each other while watching the same TV programs. Based on the choice of TV channels and conversation while watching TV, we can estimate other person’s interests and preference, which is a basis for establishing friendly relationship between others. Therefore, giving the robot a capability of recommending TV program will contribute to establish friendly communication with human partners. Furthermore, transition to digital terrestrial television broadcasting (DTTB) will bring us difficulty in finding TV program worth watching from a number of TV channels in near future. Therefore, a method for recommending TV programs will be one of the most important technologies for realizing intelligent support of our daily lives, such as a partner robot. The proposed method extracts association rules from user profiles that are generated from their TV watching logs, based on which TV programs are recommend to a target user. In order to have extensibility in terms of information resource for generating user profile, the method employs a user profile of commonly-used bookmark format. When association rules are extracted from a set of the profiles, generalization law is introduced so that variety of users’ behaviors can be reduced. Experiments are performed with actual users’ logs, and the result shows the generalization law contributes to increase the accuracy of TV program recommendation.