Personalized Movie Recommendations Based On The Hyper Graph

As a modern visual and audition art, movie contains literature, photography, music and other arts. It is a combination of modern technology and art. Recently, the rapid growth of the number of movies on the Internet provides users massive amount of resources. With the rising demand of movies, how to find the most valuable movie for one particular user is becoming a technical problem.Collaborative filtering is one of the popular methods in recommendation technologies. In the traditional recommendation based-on collaborative filtering, only the reviews of users are considered, which can lead to cold-start and sparsity problems. Movie communities like Douban and IMDB offer a wealth of information, such as tagging information and property information of movies. Effective utilization of these information can solve the problems mentioned above. Nevertheless, there are two issues to exploit this rich social media information:(1) There are many different types of objects and relations in movie social communities,which makes it difficult to develop a unified framework taking into account all objects and relations.(2) Some relationships are more complex than pairwise relation,and thus cannot be simply modeled by a graph.In this paper, we propose a personalized movie recommendation based on hyper-graph. We use hyper-graph to model the various objects and relations, and consider movie recommendation as a ranking problem on this hyper-graph. Users, movies, labels, directors and actors are regarded as vertexes of the hyper-graph and the relationships between them are regarded as edges of this hyper-graph, where the relationships between movies are obtained by items-based collaborative filtering. After that, the user as the vertexes of searching, ranking the other vertexes by manifold ranking (better connectivity will lead to higher ranking). The ranking results are based on the ranking algorithm which can realize personalized movie recommendation. Experiment on data sets in Internet Movie Database have proved that our algorithm performs better than traditional recommendation methods.