Novel automatic group identification and preference aggregation methods for group recommender systems

Group recommender systems are specialized in suggesting preferable products/services to a group of users rather than an individual by aggregating personal preferences of group members. In these systems, the initial task is to identify groups of similar users via clustering approaches, as user groups are usually not present. However, clustering users into groups commonly suffer from sparsity and complexity problems as the content in the domain proliferate. Moreover, group homogeneity and size are the critical parameters for organizing group members and enhancing their satisfaction. In this dissertation, novel automatic user grouping approaches are proposed for enhanced group formation and group size restriction. Furthermore, two different genre-based mapping strategy that transforms user ratings into a tiny and dense vector to represent users is proposed. These strategies both improve the required computation time for grouping and eliminate adverse effects of data sparsity. Finally, to construct more homogeneous group formation, two distinct strategies that utilize the genre-based similarities and the similarities based on demographic characteristics are developed. Experiments performed on two benchmark datasets demonstrate that each proposed method outperforms its traditional rival significantly. Group recommendations are commonly produced by utilizing aggregation techniques that analyze the propensities of the whole group by combining the preferences of the users in the group. Although there exist various aggregation techniques in the literature, they have some weaknesses. To overcome these weaknesses and improve the quality of group recommendations, In this dissertation, firstly, a personality-aware aggregation technique named as the PwAvg is developed. This technique determines the influence degree of each member in the group using five fundamental personality traits and then utilizes them to weight the preferences during the aggregation process. Also, to feature popular items on which group members provided a consensus, two different hybridized techniques that employ two prominent aggregation methods together in harmony is proposed. Finally, an enhanced aggregation technique built on top of the hybridized techniques and named as AwU is proposed. The AwU technique robustly considers the distribution of group members' ratings by utilizing the information entropy and consequently produces group referrals that ensure the maximum number of satisfied individuals. Experiments conducted on different datasets indicate that the proposed aggregation techniques provide more-qualified group recommendations compared to existing methods.