Leveraging item attribute popularity for group recommendation

Group Recommendation Systems (GRS) are ubiquitously available to give recommendations to users indulging in group activities. These systems recommend items based on the assumption that recommendations from like-minded users or users that rate items similarly will be ideal. However, one of the major problems faced by a GRS is the New User Problem due to the absence of any ratings from such users. In this situation, demographic filtering is exploited i.e. recommendations are predicted from ratings generated by group of users from similar demographics. It is well researched that commonly used local popularity of items results in low quality group recommendations due to inclusion of only positive ratings of the group members. Authors propose a group recommendation framework (IAPR) that leverages Item Attribute Popularity to capture overall interest of the group on items and their attributes. Valuable group recommendations for the new user are computed using a novel group aggregation strategy considering both positive and negative preferences. Experiments are conducted using Movie Lens dataset and results of IAPR are compared with two variations of IAPR and two well-known KNN based recommender systems. Results by IAPR show significant improvement in the quality of group recommendations.