Research And Implementation Of The Method Of Internet Movie Recommendation

With the development of Internet technology, Internet resources have been increasing. Faced with vast amounts of resources users are often difficult to choose. The research of personalized recommendation has been an inevitable trend. Movies resource is an important part of Internet resources. With the increasing of the number of movie videos, movies on demand simply have not met people’s needs. So personalized movie recommendation system came into being. Personalized movie recommendation system helps users quickly find movies they are interested in from a lot of movies.Personalized movie recommendation system usually consists of three parts:page display of movies in front; movie recommendation algorithm; database in back-end. The core of the system is movie recommendation algorithm. Collaborative filtering is the most widely recommendation algorithm. Against the lower accuracy of existing collaborative filtering, considering the situation of multiple users in family and the problem of cold start, this paper carried out the research of personalized movie recommendation method based on collaborative filtering. In addition, In order to improve the prediction accuracy of rating of existing recommendation, this paper also carried out the research of recommendation algorithm based on graph.The main work of this dissertation is as follows:1. Propose a movie recommendation method based on collaborative filtering, establish interest models of users, considering the idea of content filtering improve the collaborative filtering algorithm based on user, and further optimize the algorithm for sharing one information point between multiple users in family, and use a recommendation method based on similar movies to alleviate the cold-start problem. Experimental results show that improved recommendation method has higher recommendation accuracy than the existing collaborative filtering algorithm by20%.2. Propose a collaborative filtering algorithm based on graph, use the path information between item vertices in graph and consider the degree of vertices to calculate the similarity of items, and then predict item rating by the method of the collaborative filtering algorithm based on item. Experimental results show that the proposed algorithm has lower MAE values (average absolute error)than collaborative filtering algorithm based on item by0.05,and has lower MAE values than algorithm based on resource allocation by0.02.3. Design and implement an internet movie recommendation system. The recommendation system has two main functions, one is recommending personalized movie list to users under historical record of users; another is recommending similar movie list of the real-time on-demand movie.Lastly, this paper implemented a movie recommendation system. This system has been used in Internet video-on-demand, named OTT, services in Anhui Radio and Television.