Construction Of Movie Recommendation System Based On Collaborative Filtering

With the rapid development of information technology and Internet, people gradually entered into the era of information overload. In this era full of information, no matter information consumers or producers are both faced with great challenge: for information consumers, how to find the information they need from the vast amounts of information is an difficult thing; for information producers, how to show their information to the users whom they are interested in is also hard. At this time, personalized recommendation algorithm arises, it is an important tool to solve this contradiction and realize the win-win. For a recommendation system, how to make use of the user’s historical information and predict the potential content which users may be interested in is the key point.Collaborative filtering algorithms in the field of personalized recommendation is widely used. There is also the problem of sparsity in the data set, cold start problem, the accuracy issue, scalability and other issues. This paper proposes an improved algorithm for these problems based on Slope One algorithm. The improved in calculating a score value of the target item is introduced project similarity, eliminating the impact of the project on the similarity scores, while the improved algorithm to maintain the original algorithm a bit, having a data sparsity problem better adaptability.The main contributions of the work can be summarized as following:1. Build a movie recommendation system based on collaborative filtering. Movie recommendation system using the traditional B / S three-tier structure, the use of three Struts2 + Hibernate + Spring Framework, My Sql database, Tomcat server is built. The system can effectively manage users and movie-related information, can recommend movies to users which they interested in.2. Proposes an improved algorithm based on the Slope One. First, according to the historical rating data between users the similarity between items is computed. Then the nearest items are chosen to calculate the devitation between the target item and its nearest items. Finally, the similarity is used as the denominator to calculate thepredictive value of the target project.3. Experimented on Movieslens dataset, prediction accuracy, Top N and coverage are used to review the improved algorithm and comparised with the original algorithm. Experiments show that the improved algorithm not only retains the advantages of the original algorithm, but with a more high prediction accuracy, but also has good adaptability in the data sparsity problem.