Recommendation of (IP)TV Programs based on Collaborative Filtering using n-tuple Item Clustering

With the advent of multi-channel TV services, a prohibited amount of IPTV program contents becomes available to user’s sides. Furthermore, the number of IPTV service providers is rapidly increasing over internets. Therefore such information overload requires large amounts of efforts for users to search and navigate the program contents that they like to watch. In this paper, we incorporate collective intelligence by utilizing collaborative filtering (CF) for grouping community users to recommend user’s preferred (IP)TV programs so that the users can easily find and watch their preferred contents with significantly reduced efforts. The CF is performed in conjunction with a previous probabilistic framework but incorporates similarity-based grouping in terms of the attributes of IPTV programs for which each program is modeled as an n-tuple item with genre, actors/actress, channels, emission times, user’s ages and genders etc. So the item-user relevance of each user group becomes trustier. We also incorporate the time weighting factor into program preference computation which leads to reflecting timely changing user preference on (IP)TV programs. Experimental results show that proposed scheme yields 86.4% of prediction accuracy for top 5 recommended programs.