A Collaborative Filtering Recommendation Algorithm Based on Item Clustering and Smoothing

collaborative filtering is an important personalized recommendation technology in E-Commerce environment. With the expansion of the system scale, the search for nearest neighbor in the whole item space of traditional algorithm would be very time-consuming. Meanwhile, the extremely sparse user ratings data would lead to sharp decline of the recommendation system quality. Therefore, this article proposed a collaborative filtering recommendation algorithm based on item-based clustering and smoothing. Firstly, items were clustered according to the similarity by integrating user's ratings and item's attributes. Then we selected several clusters with high similarity to target item as the candidate set, from which the nearest neighbors would be found. Finally the current user rating for the active item would be predicted according to the neighbor ratings for the item. The experimental results show that the algorithm can improve the recommendation quality and enhance the system real-time performance effectively.