How to select and weight context dimensions conditions for context-aware recommendation?

Abstract Contextual information plays a key role in Context-Aware Recommender Systems (CARS). The rating prediction in CARS focuses on improving recommendation accuracy attempting to form a personalized information recommendation for users. Three key problems that affect the performances of recommender systems: (i) context condition’s selection; (ii) context condition’s weighting; and (iii) users’ context conditions matching. Context-aware approaches have the assumption that all context conditions have the same weight. These approaches ignore that users have different preferences in different contexts. To address these three problems, we introduce a novel approach for Selecting, Weighting Context Conditions (SWCC) and measuring semantic similarity between users’ situations. Evaluation experiments show that the proposed approach is outperforming the pioneering context-aware recommendation approaches of the literature.