基于项目流行度与用户行为的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on user behavior and item popularity

针对传统的基于用户的协同过滤算法在计算用户相似度时存在的问题,提出了一种改进的思路,即考虑项目的流行度和用户行为对相似度计算的影响。在此基础上设计了新的算法。通过Movie Lens数据集测试表明,新的方法能够在一定程度上提高推荐系统的性能。

To solve the problems of the traditional user-based collaborative filtering algorithm in calculating similarity of users, an improved method is presented, considering the influence of item popularity and user behavior in the similarity calculation. Based on the above ideas, a new algorithm is designed. Test through MovieLens dataset shows that the new method can improve the performance of recommendation system to a certain extent.