Movie Recommendation Algorithm Based On Improved SVD++ And User Behavior Analysis

Since the third industrial revolution,especially the emergence of the World Wide Web in 1989,the amount of information on the Internet has increased exponentially year by year.On the Internet,various forms of information are presented on web pages in different types: text,images,videos.,audio,graphics,and more.Among the many online information media,movie video has a wide audience as a content that people like to see.As an important part of cultural entertainment,film video can convey more information than words,and film and television media play a role unmatched by other media in people’s spiritual life.Therefore,for a movie video recommendation website,if a user searches for a movie of interest in the website,it takes a lot of effort to screen the video of interest,which undoubtedly reduces the user experience and is not conducive to the development of the website.Therefore,more and more websites have introduced recommendation algorithms,thereby providing users with some suggestions to present video content that may be of interest to users in front of users for screening.The traditional movie recommendation algorithm generally uses the scoring matrix to calculate the similarity between users,selects the nearest neighbor set of the target user,and then performs recommendation.However,there are widespread problems such as sparse data and low recommendation accuracy.In this paper,the traditional film recommendation algorithm is studied and improved for these problems.The main work is as follows:(1)combing the classic film recommendation algorithm and its application,as well as some scholars’ research results on the recommendation algorithm.Analyze the advantages and disadvantages of traditional movie recommendation algorithms and think about how to use the latest technology to improve the performance of recommendations.(2)At present,movie recommendation websites can be divided into two categories.The first one is only ratings and comments,but users can not tag,such as iQiyi,Tencent video,and the second is that users can tag,such as Douban movies.Aiming at the sparse problem of scoring matrix data in the first case,this paper extracts the user’s movie preference characteristics through the user’s movie review,and then constructs the preference similarity matrix,and then integrates it into the SVD++ model to improve the results of the score prediction.For the second case,the type of the movie that the user is interested in is obtained by the label of the movie that the user has watched,and the number of movie interest groups in which the user is co-located,the degree of coincidence of the user’s attention list,and the calculation among the users according to the three factors Similarity,select neighboring users,and then make further recommendations.(3)Before the film recommendation,this paper comprehensively considers the emotional tendency value of the recommended movie review,the number of comments,and the popularity of the movie,and introduces the concept of the film recommendation index,which is more objective than considering only a single factor.,improved the recommended effect.Finally,the paper validates the proposed algorithm and other algorithms on different evaluation indicators through comparison experiments.The experiment proves that the proposed film recommendation algorithm based on improved SVD++ and user behavior analysis can effectively solve the problem of data sparsity.The recommendation effect is improved to some extent.the paper validates the proposed algorithm and other algorithms on different evaluation indicators through comparison experiments.The experiment proves that the proposed film recommendation algorithm based on improved SVD++ and user behavior analysis can effectively solve the problem of data sparsity.The recommendation effect is improved to some extent.