Research And Implementation Of Personalized Library Recommendation System Based On Collaborative Filtering

With the development of the Internet,network information has increased,and the problem of "information overloading" has become increasingly prominent.An effective way to solve this problem is through search technology and personalized recommendation technology.Personalized recommendation analyzes user behavior,builds an interest model for users,predicts user behavior patterns,and then actively recommends valid information to users.Compared with search technology,it is a feature of recommendation technology that does not require users to actively type in search content.In this environment,the recommendation system is increasingly popular with network users.As one of the most popular algorithms in the recommendation algorithm,collaborative filtering algorithm has important research value and significance.Even though the application is extensive,a large number of studies have shown that the collaborative filtering algorithm still faces the following major challenges:(1)data sparseness,(2)cold start,(3)recommendation efficiency problem,(4)scalability problem.User-based collaborative filtering algorithms are widely used in recommendation systems.They analyze user behavior data,find users in areas with close interests,and recommend products with high domain user ratings to current users.The user-based collaborative filtering algorithm has high recommendation accuracy and good effect,but the data sparse problem also seriously affects its recommendation efficiency.In order to improve data sparseness and improve algorithm recommendation efficiency,this paper studies the related technology of user-based collaborative filtering algorithm and proposes an improved algorithm.The specific research work is as follows:(1)In order to alleviate the sparse data,a scoring matrix filling algorithm is proposed.The traditional algorithm fills the scoring matrix by project attribute similarity.The improved algorithm first measures the information quantity of different attribute values.The large amount of information gives the weight value when calculating the similarity.The attribute value is similar to the similarity calculation result.The impact of this is greater.This paper uses information entropy to measure the amount of information in an attribute.(2)The collaborative filtering algorithm performs calculations on the optimized scoring matrix.When calculating user similarity,the user-based collaborative filtering algorithm only calculates the similarity of user ratings.When calculating the similarity of users,this paper considers the influence of the number of public rating items of two users on the results.The larger the number of public rating items Even if the scores are different,it can be said that the two users are close to each other.In addition,when the similarity is calculated,the similarity of the user's rating times is added,and the number of items in the recommendation system is too large,and many users may have insufficient public rating items.At this time,the public scoring project can be converted into a public scoring attribute.If two users have similar scores for the same attribute,it can also indicate that the two users are close to each other.By testing on the movielens dataset,it is proved that the improved algorithm designed in this paper can effectively improve the recommendation accuracy of the recommendation system.Finally,the demand research of the book recommendation system is carried out,and the overall framework of the recommendation system is designed to ensure that the function modules and data storage can be used normally,and the improved collaborative filtering algorithm is applied.