Aggregation of preference relations to enhance the ranking quality of collaborative filtering based group recommender system

•Preference Relation based Matrix Factorization is used for Collaborative Filtering.•User and Item feature information is co-factorized in the model.•Graph aggregation with collective rationality is used for preference aggregation.•Standard ranking measures were used for evaluating the group recommendation model. The recommendation of suitable products/items for a group of users has always been a difficult task. Most of the recommender systems are designed for individual use only. However, there are many scenarios where the recommendations are intended to serve a group of users. Each member of the group has their own set of preferences, and it is challenging to satisfy each member of the group with the recommended list. It has also been observed in recent studies that mere aggregation of preferences (e.g., ratings) does not provide good group recommendations. The quality of the group recommendation depends on two essential things: the ranking quality and the aggregation strategy. The first one confirms that the higher preferred items always appear first in the list, and the second one confirms the agreement among users of the group towards the recommendation list. Hence, this study proposes a method that uses the preference relation based matrix factorization technique to obtain the predicted preference (e.g., ratings) and then uses graph aggregation strategy to aggregate the preferences of the group members. We applied collective rationality during graph aggregation to maintain consistency in preferences among group members. Three benchmark datasets were used to evaluate and compare the proposed model with other baselines in terms of ranking quality of the group recommendation.