Collaborative Filtering (CF) is one of the most popular frameworks for recommender systems. However, sparsity of user-item interactions degrades the performance of CF significantly. Using auxiliary information is a common way to solve this sparsity problem. Heterogeneous information networks (HINs), which contains a plurality of types of nodes or rich relations between nodes, make it promising to boost the performance of recommendations. In this paper, by integrating a rich variety of heterogeneous information of items into CF, we propose a novel hybrid recommendation method called Collaborative Heterogeneous Information Embedding (CHIE). CHIE jointly performs fused representation learning for items in HIN and Probabilistic Matrix Factorization (PMF), a model-based CF, for the ratings matrix. Moreover, We conduct experiments on a real movie recommendation network, which show that our approach outperforms the state-of-the-art recommendation techniques.
展开▼