Research And Implementation Of Movie Recommendation System Based On Matrix Completion And User Interest

With the rapid growth of network data in the information age,although it brings more convenient experience and superior services,it also brings the problem of information overload.How to quickly and accurately recommend for users has become a hot issue.The traditional collaborative filtering algorithm is proposed and applied in the recommendation system,which effectively improves the accuracy of the recommendation system.However,the traditional model has a series of problems.First,the lack of data in the user rating matrix leads to a low accuracy of prediction.Secondly,the increasing number of users brings the problem of computing time-consuming of collaborative filtering algorithm.Finally,for new users,the recommendation system cannot generate recommendations.However,this does not take into account the impact of time factors on user interest.To solve the above problems,this paper proposes a recommendation algorithm based on matrix completion and user interest to design and implement the movie system.The specific research contents are as follows:Firstly,in order to alleviate the problems of sparsity,scalability and similarity distortion,this paper designs the related algorithms.Its core idea is to fill the scoring matrix by using the singular value threshold algorithm(SVT algorithm)based on matrix completion through the low rank of the scoring matrix;Or use the small batch clustering method(Mini batch kmeans)to classify users,so as to ensure the normal operation of subsequent division operations.Based on the association between users and users,through the division of "nearest neighbor relationship",under certain conditions,reduce the retrieval time carried out by users in the nearest neighbor construction link;On the other hand,it is still necessary to consider the change of users’ actual interests.Based on the change of time,analyze the impact of time factor on the similarity algorithm,and use the best time factor to perform personalized recommendation operations to obtain top-N recommendation results.Secondly,for the cold start problem of the system,the recommendation information can be obtained by calculating the similarity of movies according to the category and label information of movies;On the other hand,according to the registration information of new users,the feature preference of new users will be calculated as the recommendation standard.Finally,the overall system framework is constructed,and each functional module is comprehensively designed and divided in detail.Build each module of the system through the designed functional modules,and take the performance or function as the starting point to perform the test work.After the specific test link is completed,the system can be integrated.Based on this factor,this paper considers,implements the design and realizes the control of Django,and uses the recommendation system to ensure that the subsequent verification work can be carried out smoothly,so that the system can solve the problems of cold start and sparsity in a short time.