Research On Network Structure Recommendation Algorithm

At present,the rapid development of network technology leads to the increasing amount of network data.This situation makes it difficult for people to find the goods or other information they really need when they face the vast online data.In view of this problem,for the urgent needs of users and operators,the recommender system came into being.The essence of the recommender system is to find the goods and information that they are really interested in for the user.However,the current recommender system has no way to deal with the increasing demand in terms of recommendation accuracy and personalization,and they still faced with problems such as cold start.How to solve the above problems effectively has become an important research direction of the recommender system at present.This paper based on the recommender system of graph network structure,analyzed a large number of relevant excellent literatures.Then analyzed several major problems that need to be solved in the current recommender system,emphatically improves the recommendation algorithm of graph network structure.Based on the tripartite graph network structure recommendation algorithm,introduced the tag system.By using the tags can depth to condense the characteristics of user preferences and commodity characteristics,the tags can be used as a node in the tripartite graph network structure to improve the accuracy of the recommendation algorithm.Because of the timeliness of tags,traditional recommendation algorithms and tag system can not respond well to the problem that users ' hobbies change over time.For this problem,this paper recorded the user's initial time and frequency of use the tags,and then add a time-weighted function to the tags.then proposed a timeweighted tags tripartite graph network structure recommendation algorithm.The timeweighted tags tripartite graph network structure recommendation algorithm is simulated in Movielens and Delicious data sets respectively.By analyzed the experimental results,we can verify that the recommendation algorithm proposed in this paper has higher accuracy and personalization compared with the traditional recommendation algorithm,and it can solve the problem that the user's hobby changes with time.