A Time-Topic Coupled LDA Model for IPTV User Behaviors

Internet protocol television (IPTV), the television services through the Internet, has recently gained increasing popularity. One key advantage of IPTV over the traditional terrestrial TV, satellite TV, and cable TV is that it makes the TV viewing experience more interactive and personalized. Many efforts have been made for delivering personalized IPTV services, of which understanding user preferences is the most critical. A key challenge in analyzing IPTV user behaviors is that multiple individuals may be associated with an IPTV account, thus making it difficult to identify individual users. We propose a novel time-topic coupled latent Dirichlet allocation (LDA) model, which considers the topic of TV programs viewed as well as the timestamps of the viewing behaviors, in order to capture the inherent viewing patterns of individual users along the topic as well as the time dimensions. Based on the coupled LDA model, we further summarize the behavior patterns of IPTV users. We perform an in-depth study on several intrinsic characteristics of IPTV user activities by analyzing the real-world data collected from an operational nation-wide IPTV system in China and show that the proposed coupled LDA model is able to capture interesting viewing patterns.