Personalized TV Recommendation with Mixture Probabilistic Matrix Factorization

Abstract

With the rapid development of smart TV industry, a large number of TV programs have been available for meeting various user interests, which consequently raise a great demand of building personalized TV recommender systems. Indeed, a personalized TV recommender system can greatly help users to obtain their preferred programs and assist TV and channel providers to attract more audiences. While different methods have been proposed for TV recommendations, most of them neglect the mixture of watching groups behind an individual TV. In other words, there may be different groups of audiences at different times in front of a TV. For instance, watching groups of a TV may consist of children, wife and husband, husband, wife, etc in many US household. To this end, in this paper, we propose a Mixture Probabilistic Matrix Factorization (mPMF) model to learn the program preferences of televisions, which assumes that the preference of a given television can be regarded as the mixed preference of different watching groups. Specifically, the latent vector of a television is drawn from a mixture of Gaussian and the mixture number is the estimated number of watching groups behind the television. To evaluate the proposed mPMF model, we conduct extensive experiments with many state-of-the-art baseline methods and evaluation metrics on a real-world data set. The experimental results clearly demonstrate the effectiveness of our model.

Factors Affecting the Performance of Recommender Systems in a Smart TV Environment

“…N. Chang et al [28] proposed a TV program recommender framework, which integrates the Web 2.0 features into television sets and smart TVs (set-top-boxes). A personalized TV recommendation with mixture probabilistic matrix factorization [15] developed a two-stage framework for building a TV recommender system. First, the proposed framework automatically learns the number of watching groups, and then the mixture probabilistic matrix factorization (mPMF) model was proposed for learning the mixture preference of television programs.…”

“…Therefore, the recommendations based on login information may not be relevant. Moreover, there may be diverse preferences and interests, which makes it hard to predict the exact preferences of every individual in a group [15].…”

“…However, in most households, smart TV is enjoyed by an entire family or closed group [53]. In this case, the smart TV profile (single profile) may not be the true representative of the entire group/family [15]. In content-based filtering, the already consumed contents and profile play a key role in recommendations and hence, may fail to generate relevant and better results for the smart TV watching scenario.…”

“…However, a smart-TV logged-in by a single person email address may be enjoyed by the entire group or family members, including a senior citizen, wife, husband, kids, etc. Furthermore, there may be many viewers, having diverse interests, ages, and preferences behind a single TV, which makes it difficult to predict the accurate preferences of a closed group or family [15].…”

“…The explicit feedbacks are rarely provided by a smart TV viewer. Therefore, specifically in the context of smart TV, we are left with only a few implicit feedbacks for recommending relevant items to the viewer(s) [15]. In hybrid filtering techniques, both approaches are combined to get better recommendations.…”