Recommendation algorithms for implicit information

Collaborative filtering (CF) methods are popular for recommender systems. In this paper we focus on exploring how to use implicit and hybrid information to produce efficient recommendations. We suggest a new similarity measure and rating strategy for neighborhood models, and extend original matrix factorization (MF) models to explore implicit information more efficiently. By the mean time, We extend the new MF models to integrate user or item features and obtain a new hybrid model and a corresponding algorithm. Finally we compare our new models with some well known models in our experiments.