Research On Double Neighbor Personalized Movie Recommendation Method Based On Affective Preference

With the wide application of information technology and the explosive growth and rapid development of the Internet,information resources usher in an era of high speed,diversity and intensive dissemination.But on the other hand,people have gradually entered the era of information overload from the era of information scarcity.More and more data can not be effectively used,which caused the waste of information resources.Therefore,recommendation system came into being,it's task is to contact users and information.On the one hand,recommendation system can filter useless information and help users find valuable information for themselves.On the other hand,it can make resources and information actively display in front of users who may be interested,so as to achieve win-win situation for information consumers and information producers.This paper mainly analyses several key problems in personalized recommendation system.Traditional recommendation algorithms have the problems of sparse matrix,cold start of users and projects.In the process of recommendation,traditional algorithms seldom consider the timeliness of recommendation results,ignore users' preferences for project attributes and their changing rules.In order to solve the above problems,this paper mainly studies the recommendation method combining user and project similarity double nearest neighbors,and proposes a matrix filling method based on project similarity,which solves the sparsity and synthesizes the advantages of the algorithm.In addition,on the basis of the double-nearest-neighbor recommendation results,TopN ranking of the recommendation results is carried out by adding the change of users' preferences for the attributes of the project.The main work of this paper are:(1)Establishing the recommendation method based on the idea of double-nearest neighbors.Use the binary matrix of movies to select their similar movie groups,and fill in the sparse user rating matrix.At the same time,select their similar user groups by using the filled rating matrix.Considering the similarity from the perspective of double neighbors.Make the recommendation results more accurate.(2)Calculate user preferences according to the type of movie and the year of release.This paper calculates the scores of movies with various attributes in the users' history record,and calculates user preferences based on their proportion.Take into account the user's interest in the type and the time of movies.(3)This paper considers that users' preferences will change over time,in order to reflect that the influence of users' past preferences on recommendation will reduce with the increase of time interval.User the theory of Ebinhaus forgetting curve,establish a time-effect function based on forgetting curve to calculate the dynamic change of user's interest preference.So as to make the recommendation results more time-effective.At last,mined the data of Douban Movie Network as training set and test set.The experiment result shows that compared with the traditional algorithm,the accuracy of the algorithm proposed in this paper has been significantly improved.