Diversifying Group Recommendation

Recommender-systems has been a significant research direction in both literature and practice. The core of recommender systems are the recommendation mechanisms, which suggest to a user a selected set of items supposed to match user true intent, based on existing user preferences. In some scenarios, the items to be recommended are not intended for personal use but a group of users. Group recommendation is rather more since group members have wide-ranging levels of interests and often involve conflicts. However, group recommendation endures the over-specification problem, in which the presumingly relevant items do not necessarily match true user intent. In this paper, we address the problem of diversity in group recommendation by improving the chance of returning at least one piece of information that embraces group satisfaction. We proposed a bounded algorithm that finds a subset of items with maximal group utility and maximal variety of information. Experiments on real-world rating datasets show the efficiency and effectiveness of our approach