Combining Knowledge Graph And Collaborative Filtering Book Recommendation Algorithm

With the development of technology,online bookstores have become one of the important channels for people to buy books.How to effectively recommend books to customers has become an urgent problem for book sales websites.For this reason,book recommendation algorithm came into being.The research content of this article mainly focuses on the following three parts:Book recommendation generally adopts collaborative filtering recommendation algorithm.It has the advantages of simplicity and efficiency,but it also has problems such as data sparseness,cold start,and poor interpretability of recommended results.Slope?One algorithm is an algorithm that uses a linear regression model to predict the score.Its principle is simple and easy to implement.However,the internal connection between users and items is not considered,which affects the recommendation result.A new type of data resource-knowledge graph introduces recommendation algorithms,which can optimize the quality of recommendation algorithms.In order to improve the quality of recommendation algorithms,this article proposes corresponding fusion algorithms for collaborative filtering algorithms from the perspective of auxiliary content information and fusion of semantic information.It mainly does two aspects: collaborative filtering with dynamic K-nearest neighbor Slope?One Book recommendation algorithm,a book recommendation algorithm integrating knowledge graph and collaborative filtering.First of all,in view of the problem that the Slope?One algorithm does not consider the internal connection of user books and the data sparseness of the collaborative filtering algorithm,This paper proposes a collaborative filtering book recommendation algorithm based on Slope?One algorithm incorporating dynamic K nearest neighbors.The K-nearest neighbor of the user is obtained by improving the calculation method of user similarity.K-nearest neighbor information is used to calculate the user rating deviation,the fusion item similarity is used to calculate the relevant rating information,the rating information is filled in the user-book rating matrix,and the collaborative filtering algorithm is used for recommendation on the new rating matrix.Secondly,in view of the problem that the collaborative filtering algorithm does not consider the semantic information of books,this thesis proposes a book recommendation algorithm that combines knowledge graphs and collaborative filtering.Through the knowledge graph training algorithm,the semantic information of the books is transformed into a low-dimensional vector matrix,and the semantic similarity between the books is calculated by the cosine similarity formula to obtain the semantic similarity matrix of the books,thereby obtaining the semantic neighbors of the books.At the same time,the collaborative filtering similarity calculation method is improved,and the book rating neighbors are obtained according to the book external rating matrix,and finally the rating neighbors are combined with the semantic neighbors to obtain the final book recommendation result set.The algorithm makes up for the defect that the traditional book recommendation algorithm does not consider the semantic information of the book itself,and increases the recommendation effect in the semantic layer.Research and experiments show that this article improves the algorithm by integrating the similarity of users and items into the Slope?One algorithm,and does not consider the problem of the internal connection between users and items,and improves the data sparsity of the collaborative filtering algorithm by using the Slope?One algorithm to calculate the score and fill in the score data.It does have a certain optimization effect on the book collaborative filtering algorithm.Improve the collaborative filtering algorithm by combining the knowledge graph and collaborative filtering without considering the problem of book semantic information and item cold start.The recommendation quality of the book recommendation algorithm has been improved.The two research works have certain meanings for improvement.