| With the continuous enrichment of book resources and the rapid development of modern information technology,the contradiction of “information overload” has deteriorated,causing close attention of many researchers.Currently in the academic world,search technology and personalized recommendation technology are considered to be two common effective ways to solve this problem.Compared with the search engine,the personalized recommendation technology does not rely on the user to actively input the search content.The recommendation system does not provide the user with explicit keyword guidance,and collects the user's historical behavior habit data or annotation information to build an interest model for the user.Predict whether a user likes a book and the level of preference,and then actively recommend the information that the user may be interested in to the user.In practice,the user-based collaborative filtering algorithm has a large number of applications in the recommendation system with its unique advantages,but there are still some key problems to be solved,such as data sparseness,cold start,data missing,recommendation accuracy and diversity,recommended Efficiency issues and more.Faced with the shortcomings of the traditional user-based collaborative filtering recommendation system,this paper studies the related technology of traditional user-based collaborative filtering algorithm,and proposes some improved algorithms based on this.When calculating the user similarity for the optimized user-book scoring matrix,considering the influence of the number of public rating items of two users on the result of the number of ratings,the similarity of the two users' common items is added as a threshold,and the number of user ratings Also included in the calculation process,and the number of times a user's rating of a book is converted is the number of times the user rated the attribute.The larger the number of public rating items,the more the number of ratings,even if the scores are not the same,it can be roughly considered that the interest preferences of the two users are close.In this paper,the experiment and multi-cross validation of BX-CSV-Dump dataset design are obtained in the accuracy and average absolute error of the traditional user-based collaborative filtering recommendation algorithm and the improved user-based collaborative filtering algorithm.The performance data of the aspect is shown by a graph and a table,respectively,to verify whether the user-based collaborative filtering algorithm can improve the accuracy of the recommendation system to a certain extent after the improvement. |
