Research And Implementation Of Collaborative Filtering Recommendation Algorithm Based On Matrix Factorization

With the rapid development of new technologies such as the Internet and big data,people’s lives have become more convenient,but it has also led to explosive growth of network data.In order to quickly help users find the content they are interested in,more and more researchers are working on the research of recommendation algorithms to improve the quality of recommendations and recommend content that is more in line with their preferences.However,current recommendation algorithms still have problems such as data sparsity,privacy protection and cold start,some of which are improved in this paper.Firstly,in order to alleviate the data sparsity problem and cold start problem,the hidden implicit information is introduced: user’s social information and item association information,combining these implicit information can improve the accuracy of the recommendation algorithm.This paper proposes a recommendation algorithm based on double regularized matrix decompositionMDABDRT algorithm(Matrix Decomposition Algorithm Based on Double Regular Terms).The MDABDRT algorithm adds social information regularization items for user denoising and product association regularization items incorporating user activity to the matrix decomposition model,and limits the learning of user and item latent feature vectors during the matrix decomposition process.The regular item of social information can limit the similarity between the latent feature vectors of the user and the social friends who really have the same interests and hobbies,and the item association regular item integrated with the user’s activity can limit the similarity of the latent feature vectors between the product and its related products.Thus improving the recommendation quality.On the Epinions and Ciao datasets,the performance of the MDABDRT algorithm and other algorithms on the average absolute error and root mean square error indicators are compared,and the effectiveness of the MDABDRT algorithm is verified.Secondly,since the recommendation algorithm based on double-regularized matrix factorization ignores the nonlinear relationship between user latent feature vectors and item latent feature vectors,a deep neural network is introduced to learn the nonlinear relationship between user and item latent vectors.This paper proposes a double regularized matrix decomposition recommendation algorithm based on deep neural network-DRTMDABDNN algorithm(Double Regular Term Matrix Decomposition Algorithm Based on Deep Neural Network),using neural network to discover more nonlinear hidden information between users and items,and improve The accuracy of project prediction and scoring improves the accuracy of recommendation algorithms.On the Film Trust dataset,compare the performance of the DRTMDABDNN algorithm with other algorithms in terms of mean absolute error,root mean square error,precision and recall,and verify the effectiveness of the DRTMDABDNN algorithm.Finally,a personalized shopping mall system is designed and implemented,and the MDABDRT algorithm and DRTMDABDNN model are applied to the personalized recommendation module of the shopping mall system.The system can recommend related products according to the user’s behavior and preferences,so as to achieve better personalized recommendation Effect.