Dynamic group recommender methodology: Leveraging temporal trust and confidence graphs

Group recommender systems aim to recommend items to groups with shared interests, aiming to satisfy each member. Managing trust and mutual influence within the group is a key challenge that influences the choice of items by users. These systems generate suggestions for the group based on inter-member trust. A less explored but critical aspect of this trust is its evolution, which can affect the group's item selections. This paper aims to assess the impact of time on trust in group recommendations. We begin by constructing a time-based confidence graph derived from the items selected by the group members. This graph allows us to measure the confidence levels between members and plays a crucial role in identifying their risk tolerance towards new items. Recognizing that members' risk-taking behavior can influence the group, we identify members who significantly affect group decisions. The confidence graph is periodically updated to reflect new user choices and the influence of key members. Ultimately, we introduce a novel method for calculating implicit trust based on similarity and confidence metrics, providing a recommendation list that maximizes group satisfaction based on the computed trust levels. Finally, the proposed method is evaluated using MovieLens100k, MovieLens10M, Epinions and Yelp datasets. The results demonstrate significant improvements in Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Precision, and group satisfaction measures compared to current state-of-the-art techniques.