Research On Time-aware Group Recommender System

With the rapid development of information technology,internet users are facing the problem of information overload.In order to solve this problem,scholars have launched the application of recommender system.However,in most cases,users carry out activities in the form of groups.At this time,the recommender system introduces a group recommender system.Personal recommender system with timeliness information can improve the accuracy of recommender results,but seldom consider the timeliness of group recommender.Users in the same group may have the same time work and rest behavior,and the popularity of items will change with time.To solve these problems,this paper discusses the time information and studies the influence of time on group recommender results.According to the time of users rating movies on social networking sites,this paper makes the following research.1)This paper proposes a generalized matrix factorization group recommendation algorithm based on fusion time.Firstly,the group is divided into groups.Users who have commented on the same movie in the similar time period are divided into a group,and the same user can be divided into multiple groups.Then the processed time context information is processed with the user information and movie information,and the processed results are input into the neural network for training and fitting,and multiplied by the inner product of the matrix.Finally,it improves the accuracy of project score prediction,and compares several fusion strategies,as well as the accuracy under different strategies.2)A time context group recommendation algorithm based on deep learning is proposed.The group is also divided according to time,and then the movie score is attenuated by heat attenuation function to obtain the attenuated movie score information.The group score time triplet is fitted in FM model and multilayer neural network model respectively.At the same time,low-order and high-order feature learning can improve the training speed.This paper analyzes the situation of different attenuation factors,and compares the accuracy of different algorithms.3)According to the existing recommender model,the movie recommender system is implemented.According to the recommender algorithm model,a human-computer interactive movie recommender system is built.According to the design idea of the project and the specification of software design,the movie recommender system is completed.