Design And Implementation Of E-commerce Recommender System Based On Hybrid Methods

With the rapid development of the Internet and the rapid popularization of e-commerce,people's quality of life is also improving,and their consumption habits have changed.More and more people will choose online shopping.With the increase of the number of e-commerce users,the variety and quantity of commodities are also increasing,so as to attract users and meet the diversified needs of users.Although the increase of commodities brings more choices for users,it also makes it difficult for users to quickly select the commodities they are interested in a limited time,and the problem of information overload also follows.The emergence and rapid development of recommendation system can solve this problem well.On the one hand,it can provide personalized recommendation for users,and help users to choose products,on the other hand,it can explore long tail products,and make the market more diversified.Therefore,recommendation system becomes more and more important and widely used in e-commerce.In order to further improve the recommendation efficiency and accuracy of the recommendation system,and meet the personalized needs of users for recommended products,after learning the relevant technologies and algorithms,the recommendation system is deeply understood.Based on the general recommendation system and algorithm,this paper considers and studies the problems of large data volume and data sparsity in the system and algorithm.Respectively from the perspective of commodities and users,some improvement methods are put forward.In view of the complexity of computing time caused by the large amount of commodity data,an improved collaborative filtering method based on commodity category grouping and matrix decomposition within the group is proposed,which improves the computing efficiency of the recommendation algorithm to a certain extent.At the same time,in order to effectively use the similarity between users and avoid the deviation of user neighbor selection caused by sparse scoring matrix,another collaborative filtering method based on user clustering is adopted in this paper,and the prediction results of the two recommended methods are combined by the weighted formula,and the appropriate weighting coefficient is obtained through training,and finally a commodity based method is obtained.The hybrid collaborative filtering method of category grouping and user clustering improves the overall recommendation accuracy.Then using the above hybrid recommendation method,combined with spark and other big data technologies,through the analysis of each system function and module,an e-commerce recommendation system is designed and implemented,which improves the accuracy of recommendation and algorithm efficiency.In this paper,a comprehensive functional test and performance test are carried out for the whole recommendation system,and multiple data sets are used to evaluate and analyze the recommendation effect of the system.The test results show that each functional module of the system operates normally and can meet the needs of users.Compared with a single recommendation method,the hybrid recommendation method is more stable and balanced in performance.The accuracy is also improved.By adopting spark framework and distributed computing,the recommendation efficiency and real-time performance are effectively guaranteed,and the user experience is improved.