Research Of Recommendation Algorithm Based On Weighted Tripartite Graph Model

With the rapid development of mobile Internet and the popularity of intelligent terminals,more and more consumers choose to use intelligent terminals to chat,shopping,entertainment and so on,China has become one of the most important markets of the global mobile Internet.In recent years,all kinds of App are emerging,people want to find the interested Apps from a large number of Apps by traditional search methods becomes especially difficult,so the personalized recommendation came into being.But the traditional personalized recommendation algorithm can not solve the problem of cold start,data sparsity and user interest migration very well.In order to solve these problems,This paper lead into the label information of the item,studies the personalized recommendation algorithm based on user-item-tag three element relationship.A personalized recommendation algorithm based on weighted tripartite graph model is proposed.The main research contents and achievements of this paper are as follows:1.A personalized recommendation algorithm based on weighted tripartite graph model is proposed.this paper defines the importance of user weight,weight items and tag importance weight initialization resource allocation.we introduce time parameter to reflect user interest migration by changing the weight of the three part graph.On the basis of the three plans,to improve the accuracy of the recommendation by the diffusion algorithm.In addition,the heat conduction algorithm is used to improve the diversity of the recommended objects and to balance the accuracy and diversity through the integration of resources.2.Preprocessing the real user's online traces data provided by a communication company.Firstly,processing the domain name user access into token.By defining the anti document frequency tidfof token,we can distinguish the ability of different token to express HOST.Crawling the App information of application store by crawler program,mapping the token of the domain name to App name.Then crawl the category tags in the application store according to the name of App.There are many kinds of tags that come from crawlers,and existing multiple tags with the same semantics but different text expressions,we need to summarize the tag categories and build a unified App category tag system.3.Experiments are carried out on the real data set provided by a communication company.The experimental results show that recommendation algorithm based on weighted tripartite graph model is superior to other traditional recommendation algorithms in recommendation performance.