Movie Recommendation System Based On Collaborative Filtering Algorithm

With the rapid development of IT technology,we have come to the data age,and the "data overload" problem in the data age has come unstoppable.How to find what is needed in massive data is the focus of people's concern.The successful appearance of the recommendation system has alleviated this situation.At present,the recommendation system has been well applied in e-commerce,news,music,and video websites,and has become an important tool for solving the problem of "data overload",which can help users in a huge amount of information.Search for information that interests you.However,in reality,there are still some problems.Since the recommendation system is for the user,the content of the user is found from a large amount of information,that is,the item,so the number of items and the number of users are very large,and the traditional recommendation algorithm often runs in a stand-alone environment,which makes its computing power extremely limited and cannot meet the needs of current IT services.In addition,when a new item appears in the recommendation system,there will be a cold start problem.The previous classic recommendation algorithm cannot be very Good implementation of the recommendation.How to provide recommendations quickly,efficiently,and accurately is the main research content of this paper.The specific research content is as follows:(1)The traditional object-based collaborative filtering algorithm has a cold start problem.This paper designs an improved scheme for the fusion of basic attribute similarity and movie association similarity for the cold start problem.Different similarity calculations are used for different kinds of movie attributes.The method can recommend the movie that newly enters the recommendation system to the user,effectively alleviating the cold start problem.(2)The BP neural network is introduced to implement and optimize the improved scheme in(1).Using the BP neural network self-learning and self-adaptive ability,the fusion of the basic attribute similarity and the movie association similarity of the movie is realized,and the fusion similarity is generated.Similarity is used to predict the user's rating of the movie,and the parameters are adjusted by comparison with the real score to make the recommendation more accurate.(3)The comparison shows that the optimized algorithm runs more efficiently on the Spark platform than the stand-alone operation.Therefore,based on Hadoop+Spark,a recommended system prototype system with optimized collaborative filtering algorithm as the recommendation engine is implemented.