The Research On Key Technology Of News Recommendation System Based On Hadoop

In recent years,the rapid development of Internet technology caused information overload problem to users,and make it more and more difficult for users to obtain the necessary information.Although search engine can query the desired information,but still lack inititative,so the personalized recommendation system was presented.The personalized recommendation system doesn’t need user to enter information,and able to provide the information which user interest to,so it has been widely used.Personalized recommendation system is usually deployed on big data platform to analyze and data mining,it is able to find potential user’s interest through analyze user’s login information and historical behavior and so on,and then provide personalized recommendations to improve user’s Internet experience.In the news media field,massive data is produced every day.News data tend to have distinct timeliness,classification,socialization and other characteristics.There have been some news media combine the traditional ways to push news with personalized recommendation system,which greatly enhanced the viscosity between users and news media,and make news media better into the development of Internet.Although the combination of news media and personalized recommendation system is the mainstream trend,but it still face some problems such as cold start,big data,low accuracy rate and recall rate,and so on.How to solve these problems is the key to study recommendation system.This paper combines the current research results in the recommendation system,and combines the user’s social information with personalized news recommendation system.The main achievements of this paper are as follows:1.Bundle the user’s social information and historical behavior together for recommended.2.A scalable clustering algorithm called the shortest distance clustering algorithm(SDCA)based on social network is proposed for recommendation system.This algorithm change the traditional input way of recommendation system and make the recommendation system be improved effectively.3.Combining the personalized news recommendation system with the Big Data,this recommendation system can parallel running which improve the scalability and adapts to the demand of massive news reports.4.Do experiments to the data size and related parameters,and implement a personalized news recommendation system,finally the related test such as accuracy and real-time were carried out.In this paper,Summarized the current research about personalized recommendation.Summarized the main recommendation algorithms and the problems which may face to during the progress of build and run recommendation system.Then proposed the shortest distance clustering algorithm based on social network.Finally implement a news recommendation system by Hadoop platform.