New Approaches to Mood-based Hybrid Collaborative Filtering

Recently, mood has proved to be an important contextual feature in context-aware recommender systems (CARS) by some studies. In this paper we propose two new approaches to mood-based hybrid collaborative filtering (CF) in order to further improve the performance accuracy and user satisfaction by utilizing emotional context in CARS. We first describe the traditional user-based CF as the baseline approach, and then propose a new mood-based user-based CF which detects user preferences to each emotion. On this basis, we propose two hybrid CF approaches using multiple step nearest neighbors search and predicted ratings fusion strategies respectively. We perform experimental comparisons of the above approaches on the Moviepilot dataset released for the Challenge on Context-Aware Movie Recommendation (CAMRa2010). The results suggest that both hybrid approaches provide improvements in performance.