Research On Personalized Book Recommendation Algorithm Based On Collaborative Filtering

With the rapid development of information technology and the Internet,the amount of data generated is growing exponentially,and the information overloading problem is becoming more and more serious.it is difficult for people to find what they are really interested in from the vast amount of information.Personalized recommendation technology,according to the user's age,gender and other social attributes,historical behavior and other explicit or implicit information,using a variety of recommendation algorithms to actively make recommendations for users and help users make decisions.With its strong information push ability,personalized recommendation has been widely concerned by the communication academic circles.However,the current academic research on personalized recommendation is mostly focused on its impact mechanism,and there is little introduction to personalized recommendation algorithm itself,not to mention the recommendation improvement research on optimizing its performance,so the recommendation algorithm is completely in the "black box".First of all,this paper gives a detailed text description and formula description of the main algorithms of personalized recommendation,and points out their advantages and disadvantages and applicable scene.Then,the most widely used recommendation algorithm,personalized recommendation based on collaborative filtering,is selected as the main research object,and many problems in its application performance are analyzed:First,on the calculation of user similarity,how to introduce new feature attributes to measure,so as to make users' neighbors more accurate;Second,the interference of popular items,what kind of strategies can be adopted to punish popular items;Thirdly,the sparsity of user-item scoring matrix.This paper proposes a new improvement method,including a trust model based on the community relationship,to improve the accuracy of user similarity calculation,so as to improve the effect of personalized recommendation.Finally,the improved algorithm model in this paper is tested on doubanshuishu data set,and the comparison of experimental results shows that the improved collaborative filtering algorithm has better performance than the traditional.Through the research of this paper,I hope to open the "black box" of the algorithm of personalized recommendation technology,so that the academic circles can have a clearer understanding of its recommendation principle.In addition,I hope that it can give some inspiration and provide some reference for the industry to improve the effect of personalized recommendation in the field of news.