Research On Recommendation System Based On Forget Theory And Weighted Bipartite Graph

Recently, the electronic commerce and network technology develop rapidly,Smartphone 、tablet PCs and other portable devices rapidly growing popularity, A large amount of information is uploaded to the Internet, The size of the Internet and Information continues to expand, the information of the world on the state of the Big Bang,people goes from lack of information era into the era of information overload,In this day and age,The amount of information which is poor quality and low value and people face them every day,As a custom of using information,How to find the interested information from a large amount of information is a very difficult thing,At the same time,As information producers,How to make their own production of information stand out,by the majority of the user’s attention is a very difficult thing.In order for a user to convenient,accurate to obtain information,solve the problem of information overload,recommender systems emerge as the times require.Recommendation System can obtain user behavior data active or passive,such as user’s browsing history,scoring information,keyword search behavior information.On the one hand,Recommend system predict user preferences,to help users find the information which they interest.On the other hand,let the information can be displayed in front of the user who interested,in order to achieve “win-win” for both sides.Currently, there are four main recommendation algorithm:Content-based recommendation algorithm,Collaborative filtering recommendation algorithm,graph structure recommendation algorithm and Hybrid recommendation algorithm.there are many other recommendation algorithm,such as association rules recommendation algorithm,Knowledge-based recommendation algorithm and so on.But there are many problems in the current recommendation algorithm,such as recommendation accuracy,scalability,data sparsity and cold start,user interest drift.In this article,we study on reducing the impact of recommendation precision case by user interest drift and solving cold start problem.Aiming at the problem of user interest drift,This paper presents an improved recommendation algorithm:recommendation algorithm base on Forgotten theory and weighted bipartite graph,Aiming at the problem of cold start,This paper combines the result of Content-based recommendation algorithm,Aiming at the problem of the efficiencyof recommendation system,this paper transplanted the recommendation algorithm to the Spark cluster, improve the recommendation algorithm execution efficiency.This paper firstly introduces the background of related research and research status at home and abroad, and then gives the main research of this paper work, put forward the proposed improved algorithm, then evaluates the proposed improved algorithm and compared with the existing recommendation algorithm, finally carries on the summary to this article.