A Recommendation Algorithm Combining User Grade-Based Collaborative Filtering and Probabilistic Relational Models
Ying Gao, Hong Qi, Jie Liu, Dayou Liu · 2007
Collaborative filtering (CF) is a successful technology for building recommender systems. Unfortunately, it suffers from three limitations - sparsity, scalability and cold start problem. To address these problems, a recommendation algorithm combining user grade-based collaborative filtering and probabilistic relational models (UGCF-PRM) is presented. UGCF-PRM integrates user information, item information and user-item rating data, and uses an adaptive recommendation strategy for each user. In UGCF-PRM a user grade function is defined and a collaborative filtering based on this function is used, which can find neighbors for the target user efficiently. Because of the first-order character of probabilistic relational models, UGCF-PRM can solve the cold start problem. The experiment results on the MovieLens data set show that UGCF-PRM performs better than a pure CF approach in both recommendation quality and recommendation efficiency.