Hybrid recommender system with core users selection

Recommender system plays an increasingly important role in identifying the individual’s preference and accordingly makes a personalized recommendation. Matrix factorization is currently most popular model-based collaborative filtering (CF) method that achieves high recommendation accuracy. However, similarity computation hinders development of CF-based systems. Preference obtained only depends on explicit rating without considering implicit content feature, which root cause bias. In this paper, feature items described by fuzzy sets integrated into computation, helps to improve accuracy user modeling. The importance then defined according preferences, serves as baseline standards core users selection. Furthermore, based matrix model (CU-FHR) established, genetic algorithm used predict missing items. Finally, MovieLens test performance our proposed method. Experiments show CU-FHR better prediction compared with other methods.