<p>Group recommendation systems support collaborative decision-making by analyzing interaction patterns among multiple users and are employed across diverse application domains. They facilitate decision-making for a group of users by leveraging the individual preferences of its members. Accurate prediction of group–item interactions requires modeling the influence of individual members in decision-making alongside user-level and group-level preferences. To this end, we propose an influence-aware graph neural collaborative modeling for group recommendation that models both the individual user’s influence and the influence of items based on the users who have interacted with them. We employ the concept of masking to learn the influence factors of individual nodes in interaction prediction. This approach learns a soft mask over all individual nodes of an interaction graph to determine the influencing nodes. Our model captures group–item interactions by using influence-aware node embeddings for both groups of users and items, thereby improving overall recommendation accuracy. We evaluate the proposed approach on two real-world datasets, CAMRa2011 and Mafengwo, and demonstrate consistent improvements over state-of-the-art baselines on the standard evaluation metrics.</p>