Building an IPTV VoD recommender system: An experience report

Internet Protocol Television (IPTV) is an increasingly popular multimedia service which is used to deliver television, video, audio and other interactive content over proprietary IP-based networks. Video on Demand (VoD) is one of the most popular IPTV services, and is very important for IPTV providers since it represents the second most important revenue stream after monthly subscriptions. In addition to high-quality VoD content, profitable VoD service provisioning requires an enhanced content accessibility to greatly improve end-user experience. Moreover, it is imperative to offer innovative features to attract new customers and retain existing ones. To achieve this goal, IPTV systems typically employ VoD recommendation engines to offer personalized lists of VoD items that are potentially interesting to a user from a large amount of available titles. In practice, a good recommendation engine does not offer popular and well-known titles, but is rather able to identify interesting among less popular items which would otherwise be hard to find. In this paper we report our experience in building a VoD recommendation system. The presented evaluation shows that our recommendation system is able to recommend less popular items while operating under a high load of end-user requests.