Aggregation strategies in user-to-user reciprocal recommender systems

Reciprocal Recommender Systems are a class of Recommender System for recommending people to people, where the preference relationship is bidirectional and both sides must be considered in the recommendation process. Reciprocal recommender systems often generate preference values using similar algorithms to traditional recommender systems, and then they combine two preference scores into a single indicator of the compatability between the two users. None of the few existing approaches for reciprocal recommendation have investigated the influence of different aggregation functions in these systems as of yet. Our work concentrates on exploring diverse approaches for aggregating bidirectional preference relations, and presents the results of a comparative study.