Time-Aware Collaborative Filtering for Recommender Systems

Traditional collaborative filtering algorithms only take into account the users' historical ratings, which ignore the user-interest drifting and item-popularity changing over a long period of time. Aiming to the above problems, a time-aware collaborative filtering algorithm is proposed, which tracks user interests and item popularity over time. We extend the widely used neighborhood based algorithms by incorporating two kinds of temporal information and develop an improved algorithm for making timely recommendations. Experimental results show that the proposed approach can efficiently improve the accuracy of the prediction.