Research On A Personalized Recommendation System Based On Web Mining

With the rapid development of the Internet,we move into the era of big data.The information-overloading problem is becoming increasingly obvious than ever before.Recommendation system arises at this historic moment,which sets up communication bridge between producer and consumer.It pushes personalized information to users according to user preferences.This paper introduces the personalized recommendation technology,through the analysis of the principle of collaborative filtering algorithm and its shortcomings,improves the traditional collaborative filtering recommendation algorithm.On this basis,a personalized recommendation system based on Web mining is constructed.This paper studies and designs a personalized recommendation system based on Web mining.Mainly completed the following research work:1.The personalized information push method based on collaborative filtering algorithm is studied,which solves the problems of the traditional cooperative filtering,such as the neighbor set misjudgment,the interest elegant and so on.2.Optimize the similarity calculation method,considering user difference and item diversity.3.Simulate the human brain memory dynamically change user interest,considering the time dimension on user interest.4.Design the recommendation credibility mechanism to predict the user's score on the item to improve the algorithm in three aspects,improve the accuracy of recommendation.5.Aiming at the problems of data sparsity and cold-start in traditional collaborative filtering recommendation algorithm,this paper introduces Web Mining technology,build a new user interest model with comprehensive utilization of user's implicit and explicit behavior.By mining the user's behavior logs,the user's implicit behavior is converted into an implicit user item score,and an explicit user item score is used to fill the explicit user item score.The experimental results show that the personalized recommendation system based on Web mining can further improve the recommendation accuracy and coverage.