An enhanced factorized model based on user and item features
Chuanfei Luo, Yihao Zhang, Weiyao Lin, Yulin Wang, Weijie Yu · 2014
Nowadays, people rely on the Recommender System (RS) to make their decisions on the Internet. Most previous techniques only focused on the user bias and item bias, while the detailed features of users or items have been overlooked. In this paper, we propose a new algorithm to provide more precise suggestions for users in RS. The proposed algorithm introduces a User-Item-Feature-based (UIF) model to dig out concealed factors that influence the users' rating preferences. With this method, more detailed information, including the users' gender, age and occupation features and the items' decade features, can be well considered when recommending to users. Experimental results demonstrate that by adding the user and items' features, our proposed UIF model can provide more accurate and scalable results compared with previous methods.