TV Recommender System with Pattern-Based Clustering

In this thesis, we present a novel framework and an algorithm for digital TV users with the program recommendation on a different navigation interface. Traditionally, users often watch TV passively. However, we provide a concept of watching TV on user's own initiative and on his/her demand. We apply a pattern-based clustering algorithm for collaborative recommendation on back-end computation. Compared with traditional clustering algorithms which cluster points base on distance, our algorithm clusters all connotative patterns of each transaction. Hence, our algorithm is suitable for Digital TV recommender system. In addition, our system is built on MHP platform of Interactive Broadcast Profile to provide personalized electronic program guide and support the clients without return channel. Therefore, this system is suitable for the people's demand in the coming era of Digital TV.