Design And Implementation Of Film Recommendation System Based On User Behavior

With the rapid development of the Internet,people are becoming more and more dependent on the Internet in every aspect of their lives.More and more Internet users habitually get the data they need from the Internet to solve many problems in their lives.But the vast amount of data mixed up on the Internet has made it difficult for people to identify the data they need.In order to solve this contradiction,personalized recommendation technology comes into being.As the core of the personalized recommendation system,the recommendation algorithm has always been the focus of attention and research.Among the many recommended algorithms,the most widely used is the collaborative filtering algorithm.Traditional collaborative filtering algorithm using Internet users input to score as the algorithm of the project,ignoring the user behavior(purchase,collection,forwarding,project clicks)to the project scale and the effect of causing the project evaluation not accurate enough,users-project matrix is too sparse,eventually lead to recommend effect precision is low.In order to improve the efficiency of the personalized recommendation technology,this paper proposes a clustering recommendation algorithm based on user behavior,and on the basis of the algorithm,and implemented a personalized movie recommendation system based on B/S architecture.The algorithm is improved from two aspects: one is the comprehensive consideration of user behavior information and project category preference information.Firstly,the user's web log is obtained by realizing the third party interface provided by the developer,and then the user's behavior information is extracted.According to the experience,different user behaviors are given different linear weights,and the user behavior is converted into the score of the project to fill the sparse user-item scoring matrix.The population-based user-project scoring matrix has a certain degree of decline in data sparsity.Secondly,according to the population-based user-project scoring matrix,k-means clustering is carried out according to the project category,and multiple clustering clusters are generated.Calculate the distance between the target user and the cluster cluster and classify the target user into the nearest cluster.Finally,in the cluster class of the target users,the traditional user-based collaborative filtering algorithm is used to generate personalized recommendation lists for target users.Second,considering the traditional cosine similarity calculation is not sensitive to distance,this paper proposes a method to calculate the weight of the fusion score.When calculating the nearest neighbor of the clustering of target users,the accuracy of nearest neighbor calculation is improved and the recommended accuracy is improved.This paper adopted a system of douban film for third-party developers to provide the interface to get related user behavior data,using the comprehensive F test(F-test)and accuracy(Precision)to measure the quality of the recommendation algorithm,by comparing the traditional user-based collaborative filtering(CF)algorithm and the recommendation algorithm based on user clustering(the UCCF),prove that the proposed clustering recommendation algorithm based on user behavior(ABUCCF)is effective.