An Improved Collaborative Filtering Approach Based on User Ranking and Item Clustering

Collaborative filtering is one of the most successful technologies applied in recommender systems in multiple domains. With the increasing growth of users and items involved in recommender systems, some inherent weaknesses of traditional collaborating filtering such as ratings data sparsity, new user problems become more and more manifest. We believe that one of the most important sources of these problems is the deficiency of user similarities based on all users and items in authenticity and accuracy. In this paper, we propose an improved collaborative filtering method based on user ranking and item clustering, in which the users are classified and ranked in multiple item clusters by computing their rating qualities based on the previous rating records, and items are recommended for target users according to their similar users with high-ranks in different item categories. Experiments on real world data sets have demonstrated the effectiveness of our approach.