Improved personalized recommendation based on user attributes clustering and score matrix filling

Abstract Personalized Recommender Systems (RS) are used to help people reduce the amount of time they spend to find items they are interested in. Collaborative Filtering (CF) is one of the most successful techniques in RS. Data sparsity makes the result of recommendation inaccurate. In this paper, a novel evolutionary clustering method is presented. The goal of our algorithm is to gather users with similar interest into the same cluster and recommend items for users that they might like. Firstly, user attribute distance is calculated. According to the constructed network clustering model, states of users evolve over time. States of users would be stable after some period of iteration. In light of stable states of users, they are clustered into several groups. The user’s interest preferences change over time, while the user’s interests are relatively stable over a shorter period of time. Secondly, we fill the rating matrix based on scoring time and item genres. Thirdly, user-based collaborative filtering is adopted in each cluster. Similarities between users only in same cluster are computed with the filled matrix. Finally, the target rating is calculated according to the neighbor set of the users, and the top-N interested items are recommended to the target user. Through experiments, compared with the existing recommendation algorithms, the proposed algorithm improves the precision of recommendation and solves the sparse problem of rating matrix effectively.