The Research On Algorithm Of Context Tensor Decomposition Recommendation Based On User Nearest Neighbor

The rapid development of the computer technology has brought a serious problem of information overload,which virtually increase the difficulty of users to obtain the information they want.The personalized recommendation system is produced in such a situation,which records the user's historical interaction behavior and analyzes them in detail.Based on the results of these analyzes,it can recommend the products for users who may be interested in.Personalized recommendation system not only provides better user experience for users,but also personalized decision-making mechanism for the website itself.This paper introduces the related concepts involved in the personalized recommendation system,the classical recommendation algorithm and the application scenario of the recommendation system.The recommendation algorithm of two-dimensional matrix decomposition and the recommendation algorithm of high-dimensional tensor decomposition are mainly studied.In this paper,some corresponding improvement schemes are proposed for data sparseness and cold start problems of those algorithms.The main works of this paper are as follows:(1)Aiming at the data sparsity problem existing in the traditional matrix decomposition collaborative filtering algorithm,a matrix decomposition recommendation model integrating social information is proposed.It is often found that suggestions from our friends affect our buying behavior in a subtle way.The buying behavior of users is not only related to their own interests,but also influenced by the friends they trust in.Based on the traditional SVD decomposition model,this paper considers the influence of the inherent attributes of users and make use of the friendship in social network to modify the matrix decomposition model,and then use random gradient descent method to decompose the matrix.The experimental results show that the improved algorithm has better recommendation than traditional SVD recommendation algorithm.(2)Aiming at the problem of the accuracy of the recommendation algorithm based on tensor decomposition,an N-dimensional tensor decomposition algorithm is proposed.Firstly,the context-sensitive information is introduced,and the implicit feedback information in the context perception is used as the third dimension of the tensor toestablish the N-dimensional tensor decomposition model.At the same time,in order to improve the recommended quality further,the neighbor information of users is introduced to optimize the N-dimensional tensor decomposition Model,which improves the accuracy of tensor decomposition recommendation algorithm.The results show that the tensor decomposition recommendation algorithm combined with user nearest neighbor has better accuracy than the traditional tensor decomposition algorithm,and can effectively solve the problem of sparseness and accuracy.(3)The matrix decomposition recommendation model of fusion social and neighborhood information proposed and the N-dimensional tensor decomposition algorithm of fusion user's neighbor informationThe matrix decomposition recommendation model integrating social information and the N dimensional tensor decomposition algorithm integrating user neighbor information are applied to the soft-installed electronic commerce system.Besides introduce the overall architecture and technical selection of the recommendation system.In order to meeting multiple recommendation some classic recommended algorithms are used.The system is mainly divided into data layer,recommended algorithm layer,application interface layer and application layer.The system recommended modules include "read and see","buy and buy","guess you like","with recommended" and so on.The recommended system for electricity supplier system not only brings a better user experience,but also attracted many users.