Mobile-to-Mobile Video Recommendation
Mobile device users can now easily capture and socially share video clips in a timely manner by uploading them wirelessly to a server. When attending crowded events, such as an exhibition or the Olympic Games, however, timely sharing of videos becomes difficult due to choking bandwidth in the network infrastructure, preventing like-minded attendees from easily sharing videos with each other through a server. One solution to alleviate this problem is to use direct device-to-device communication to share videos among nearby attendees. Contact capacity between two devices, however, is limited, and thus a recommendation algorithm, such as collaborative filtering, is needed to select and transmit only videos of potential interest to an attendee. In this paper, we address the question: which video clip should be transmitted to which user. We proposed an video transmission scheduling algorithm, called CoFiGel, that runs in a distributed manner and aims to improve both the prediction coverage and precision of the collaborative filtering algorithm. At each device, CoFiGel transmits the video that would increase the estimated number of positive user-video ratings the most if this video is transferred to the destination device. We evaluated CoFiGel using real-world traces and show that substantial improvement can be achieved compared to baseline schemes that do not consider rating or contact history.