Research on Similarity Metrics for Collaborative Filtering

Nowadays,the recommendation system has been paid more and more attention and is more and more popular.The collaborative filtering algorithm is the most widely used personalized recommendation technology.After simply elaborating the user-based and the item-based collaborative filtering recommendation algorithms,this paper focuses on the similarity metrics including correlation similarity,cosine similarity and adjusted cosine similarity.Then,in the case of sparse data,it analyzes and compares these three similarity metrics.And the last,this paper comes to the final conclusion and proposes an improved similarity calculation method.