Declarative User-Item Profiling Based Context-Aware Recommendation

Context-aware recommendation has attracted much attention due to its ability to effectively finding the items that a target likes out of an abundance of online items. Different users may characterize different contexts to items since they also consider different contexts when they select items. Comprehensive identification of the declarative dominant contexts for both items and users can significantly affect the quality of the recommendation, which is often overlooked by the existing research. In this paper, we propose a new recommendation approach, which identifies the dominant contexts as declared by users on their previous transactions. Firstly, we identify the significant contexts from both item and user perspectives and construct the user-item profile in a personalized manner. Secondly, we propose a new context-aware recommendation model that seamlessly incorporates both declarative profiles into the recommendations. Finally, we demonstrate the effectiveness of the proposed method by conducting comprehensive experiments over two real benchmark datasets. The experimental results show that the proposed method outperforms the state-of-the-art methods.