Multi-Objective Optimization Based Group Recommender System

Abstract—Most existing studies in Group Recommender Sys-tems (GRS) explore fairness and diversity independently. It is challenging to consider the feasibility of simultaneously incorpo-rating fairness and diversity in Group Recommender Systems. This paper examines their quantitative relationship and pro-poses a framework that integrates fairness and diversity within the recommendation process. The method clusters users into subgroups based on their preferences, creates pseudo-users to represent these subgroups, and aggregates the recommendations to form a final group recommendation list. The aim is to enhance the fairness and diversity of recommendations. This design aims to balance user-level fairness and diversity while maintaining accuracy. Experimental results on the MovieLens dataset show that the proposed method outperforms the baseline method across different evaluation metrics, demonstrating the importance of simultaneously optimizing fairness and diversity in group recommender systems.