Community-based Movie Recommendation via Multimodal Network Learning
With the rapid development of Internet movie industry, Community-based Movie Recommender system (CMR) has become a popular online web service, which provides relevant movies to the targeted users. In order to provide high relevant movies, many existing movie recommendation approaches learn the user ranking model from their relative preference feedback and movie contents, which suffer from the sparsity problem of CMR data. In this paper, we consider the problem of community-based movie recommendation from the viewpoint of learning multimodal ranking metric heterogeneous network representation. We propose the heterogeneous CMR network that exploits movies' multimodal contents including textual description and visual poster, users' relative preference and their social relations for movie recommendation. We then present a novel multimodal ranking metric network learning framework, named as MRMNL, such that the learned multimodal ranking metric is implicitly embedded in the heterogeneous network representation for recommendation. We develop a random-walk based learning method with multimodal neural networks for learning ranking metric heterogeneous network representation. The extensive experiments on a large-scale dataset from a real world community-based movie recommender site show that our method achieves better performance than other state-of-the-art solutions to the problem.