Design And Implementation Of Distributed Recommendation System Based On Improved Collaborative Filtering Algorith

The high-speed growth of information data has brought about information overload.How to dig out the potential value of data in massive data and make better recommendations for users is related to the survival and development of an enterprise.The recommendation effect of recommendation system is inseparable from the support of recommendation algorithm and recommendation model.Collaborative filtering algorithm as the mainstream recommendation algorithm plays a very important role in providing better recommendation services.Therefore,this paper improves the traditional collaborative filtering algorithm and designs and implements a distributed movie recommendation system on this basis.(1)Research on recommendation algorithm based on collaborative filtering.Aiming at data sparsity and single user rating similarity,low precision similarity measurement reduces the performance of recommendation system.A collaborative filtering algorithm based on project attributes and user interest migration is proposed.Firstly,the user interest is no longer defined only by the user’s rating of the project,but by considering the project category attribute to reflect the user interest,and the project attribute similarity calculation method is constructed;Secondly,considering the user’s interest will occur dynamic migration,the introduction of Ebinhaus forgetting curve,build user interest calculation method,quantitative user interest change;finally,a new similarity calculation model is constructed.For the traditional collaborative filtering algorithm,due to the lack of user history rating data,it is unable to make accurate recommendations for users.The addition of new users or new projects still has the problem of cold start.A collaborative filtering algorithm based on matrix decomposition and clustering is proposed.On the basis of matrix decomposition,the project attributes are clustered,and the user prediction score matrix obtained by multiplying the updated project feature matrix and user feature matrix is used for personalized recommendation.(2)The analysis and design of distributed recommendation system based on collaborative filtering algorithm.On the basis of the proposed improved collaborative filtering recommendation algorithm,according to the software engineering design idea,the demand analysis of the distributed movie recommendation system based on collaborative filtering algorithm is carried out from the aspects of technical feasibility,economic feasibility,function and non-function.It is clear that the system adopts the B/S three-tier architecture,expounds the distributed design of the system,and designs each functional module in detail.(3)Implementation and test of distributed recommendation system based on collaborative filtering algorithm.The distributed movie recommendation system based on collaborative filtering algorithm is based on big data platform.The data is stored in ES and document database Mongo DB.The non-relational database Redis realizes high-speed reading of data,and combines content-based,real-time and offline non-personalized recommendation methods for hybrid recommendation.In order to verify whether the system meets the development requirements,the system function test and performance test of concurrent access in distributed environment are carried out.