Realtime Channel Recommendation: Switch Smartly While Watching TV

Existing works on TV recommendation mostly focus on determining users preferences of TV shows. A realistic system should also consider the dynamics of the shows information. In this paper, we consider the profit maximization problem for real-time channel recommendation: given a specified user and a time window, a recommender algorithm is required to decide when and how to switch among n channels, each of which contains at most k live shows. The objective is to maximize the users overall profit, i.e., the total score via watching shows minus the total cost by switching among channels. For the offline version, an exact algorithm with the time complexity \(O(kn^2)\) is proposed, a lower bound \(\varOmega (n \log n)\) of the time complexity is given for any exact algorithm. The online version is also studied. For both the non-restricted and the restricted variants, algorithms with running time \(O(n \log n)\) and constant competitive ratios are presented respectively.