Building Recommender System for Media with High Content Update Rate

A wide variety of recommender systems have been deployed on different media websites. Nonetheless, providing personalized recommendations for hundreds of new users and articles appearing daily still appears to be a challenge. In this paper, we propose a method for generating personalized recommendations for media resources with high content creation rate. We explore two general approaches, gradient boosting model and matrix factorization algorithm. Preliminary experiments showed that gradient boosting performs better for our website in terms of training speed, resource consumption, and scalability.