Abstract: As a successful technology used in the recommender system,collaborative filtering has been widly concerned by scholars in various fields.However,with the increasing of new users and items,collaborative recommendation is facing serious challenge of “cold start”.This study measured the recommending ability of user based on popularity and long-tailed distribution,and then constructed a global core user set for recommadition using user popularity,which can be used to solve “cold start” problems in recommendation systems.In additional,experimental results show that the core use set used for collaborative recommending can reduce complexity of looking for similar users without lowing the recommendation performance.So it also can be used to improve real-time recommendation.

Key words: Collaborative filtering,Core users,Long-tailed distribution,User popularity