Preference Learning Based on Dynamic Dual-Cognition for Group Recommendation

Chunlong Wang, Yue Kou, Derong Shen, Xiangmin Zhou, Tiezheng Nie, Ge Yu, Dong Li

Abstract: Group recommendation aims to suggest items that cater to the preferences of all members within the group. However, existing models often overlook the dynamic cognitive changes of group members, leading to inaccuracies. In this paper, we propose a novel Preference Learning framework based on Dynamic Dual-cognition for Group Recommendation (PL-DDGR), which aims to enhance group recommendation accuracy. Specifically, we first propose a preference learning approach based on graduality cognition, which can better understand and predict the subtle yet continuous shifts in member preferences. Then we propose a preference learning approach based on conformity cognition, which can capture the evolving nature of member conformity. We also propose a self-supervised multi-task joint training mechanism to optimize the learning of both graduality and conformity cognition simultaneously. The experiments demonstrate the effectiveness and superiority of our proposed framework.