A context-aware recommendation system considering both user preferences and learned behavior

Context awareness is an important aspect to be considered for intelligent recommendation systems. In this paper we consider the TV recommendation scenario. We argue that content-based recommendation is best suited for an environment where a database of similar user's ratings of the program is not available. Also, it is important to consider both user preferences as well as learned user behavior. We present our design of a context-aware TV recommender which considers the user's context, user's preferences and user's TV viewing behavior. We consider an algorithm reported in the literature which uses user modeling to learn user's viewing habits, and extend this algorithm to add context-based and user preference-based recommendations. Finally, we present our results of a user study which validates the efficiency of our algorithm.