A Self-Adaptive Context-Aware Group Recommender System

The importance role of contextual information on users’ daily decisions led to develop the new generation of recommender systems called Context-Aware Recommender Systems (CARSs). Dependency of users preferences on the context of entities (e.g., restaurant, road, weather) in a dynamic domain, make the recommendation arduous to properly meet the users preferences and gain high level of users’ satisfaction degree, especially in a group recommendation, in which several users need to take a joint decision. In these scenarios may also happen that some users have more weight/importance in the decision process. We propose a self-adaptive CARS (SaCARS) that provides fair services to a group of users who have different importance levels within their group Such services are recommended based on the conditional and qualitative preferences of the users that may change over time based on the different importance levels of the users in the group, on the context of the users, and the context of all the associated entities (e.g., restaurant, weather, other users) in the problem domain. In our framework we model users’ preferences via conditional preference networks (CP-nets) and Time, we adapt Hyperspace Analogue to Context (HAC) model to handle the multi-dimensional context into the system, and sequential voting rule is used to aggregate users’ preferences. We also evaluate the approach experimentally on a real-word scenario. Results show that it is promising.