A Collaborative Filtering Algorithm based on Improved Similarity

The collaborative filtering recommendation implements recommendation by using his neighbor user's preference. And the similarity calculation is the key. The traditional similarity calculation neglects the impact of co-rating item number and user average rating on similarity calculation. This causes the poor similarity calculation of users in case of sparse data. This paper introduces the two improved factors to the improved algorithm, so as to improve the traditional similarity. Meanwhile, the improved recommendation algorithm has been applied to film recommendation system. The simulation experiment shows that the improved recommendation algorithm can get a lower MAE value than traditional recommendation algorithm. In addition, the improved algorithm can improve the quality of film recommendation system.