An implicit feedback integrated LDA-based topic model for IPTV program recommendation

Abstract

Internet protocol television (IPTV), the television services through the Internet, has become more and more popular in recent years. Many recommendation systems have been made for delivering personalized IPTV services, of which understanding users' preference is the most critical. The traditional recommendation system only considers the users' playing behavior, but other implicit feedback behaviors of users, such as browsing, collecting also reflect the users' preference. We propose a novel latent Dirichlet allocation (LDA)-based model, which considers users' playing behavior as well as the implicit feedback of browsing and collecting, to capture the inherent viewing preference of individual users. The implicit feedback integrated LDA model employs three LDA models (the playing, browsing, and collecting behavior topic model), which are integrated via TV program characteristic. Based on this, we further calculate the ratio of each behavior to the recommended results by logistic regression algorithm. The experimental results show that the proposed topic model yields an average 32.5% precision for recommending 10 videos and 200 topics in IPTV recommendation, and its performance is an average of 19.5% higher than that of LDA using playing behavior only.