A study on automatic TV program recommendation based on collaborative filtering for IPTV personalization

Abstract

Due to the rapid increase of available contents via the convergence of broad-casting and internet, the efficient access to personally preferred contents has become an important issue. In this thesis, for recommendation using collaborative filtering technique is studied. For recommendation of user preferred TV programs, our proposed recommendation scheme consists of offline and online computation. At first during offline computation, (1) reasoning each user``s preference in TV programs in terms of program contents, genres and channels; and (2) clustering users based on each user``s preference profile by dynamic fuzzy clustering method. And secondly during online computation after active user logs in, (1) pulling similar users to an active user by similarity measure based on the standard preference list of active user; (2) sorting the watched TV programs of the similar users according to the preferences; (3) filtering-out of the sorted TV programs which do not exist in EPG; and (4) ranking of the remaining TV programs for recommendation. Especially, in this thesis, the BM (Best Match) algorithm is extended to make the TV programs be ranked by taking into account the TV program preference. For the proposed scheme for collaborative filtering based TV program recommendation, the proposed extension to the BM model for ranking outperforms the $\It{J. Wang}$``s rank model. The experimental results show that the proposed scheme with the extended BM model yields for the 5 item recommendation, 79% in prediction accuracy for which the $\It{J. Wang}$``s rank model resulted in 54% in prediction accuracy with 109 people.