Social- and Content-Aware Prediction for Video Content Delivery

An ever-increasing number of videos in mobile social networks increase the waiting time of video downloading. Video prefetching is a viable way to save time during video downloading, while prefetching all the videos is resource wasteful if the videos are not watched. Therefore, careful prediction of whether a user will watch a video is critical for efficient video content delivery. Most existing work on social media recommendation focuses on the top-N problem to recommend multiple videos. In this paper, we deal with the problem of predicting whether a video will be watched by a user for efficient video content delivery in mobile social networks. We propose a Social- and Content-aware Video content delivery Prediction method (SCVP) for the problem by capturing the intrinsic relationship among users and videos. We design five metrics to estimate the factors of active degree of users, social tier between users, similarity between videos, similarity between user interest and video content, and video popularity. We then use combined prediction for the video content delivery prediction to incorporate the impacts of the five factors on the prediction. Finally, we conduct experiments through simulations. Experimental results demonstrate that the proposed method SCVP can predict whether a video is watched by users with high accuracy.