Research On Personalized Recommendation Algorithms Of Web Book Resources

With the rapid development of the Internet and Big Data technology,there are more and more online book resources.There is a growing demand from users who are interested in books,but it is difficult to find books for users among the amount of book resources.In order to help users find the books that they like quickly,the personalized recommendation algorithm is applied in the field of the web library resources.Content-based recommendation algorithm and neighborhood-based recommendation algorithm are most widely used recommendation algorithms.This paper proposes a collaborative filtering recommendation method which combines the advantages of these two methods.We collect the Data of the Douban Reading Platform analyze the similarity of users and books and use the Doc2 Vec method,which significantly improves accuracy and recall rate.However,to find a book that the user likes,the recall rate decreases quickly as the number of recommendations becomes smaller.In order to improve this problem,this paper proposes to use the machine learning method to order the results of collaborative filtering algorithm and recommend the top 10 or more books to users.Firstly,we construct the features from the user behavior data,which contains user behavior characteristics,book behavior characteristics and user-book interaction behavior characteristics.The constructed features are substituted into the logistic regression and gradient boosted decision tree,and we predict recommendations.We find that the machine learning method does improve the recall rate of the recommended algorithm.Finally,this paper constructs a recommendation index to improve the diversity of the recommendation list.The empirical result shows that the diversity of recommendation results is improved with the least sacrifice of accuracy.