Application Of Deep Neural Network In Personalized Recommendation Of Book Borrowing

With the application of new information technology represented by big data and artificial intelligence new technology,information technology promotes the continuous development of business information.Modern library basically bid farewell to the traditional means of manual registration and borrowing before,and applied the book information system to carry out the related business of University book borrowing and returning,so the readers can query and borrow books more conveniently.With the continuous expansion of the number of books in the university library,readers can't find the books they need effectively in the face of massive book data,so they urgently need a means to help readers and books establish an efficient connection.Compared with other means,recommendation system is a more effective means.The recommendation system first collects the historical borrowing behavior data of readers,and then analyzes them to recommend the books that readers are interested in.The traditional recommendation algorithm represented by collaborative filtering algorithm is the most classical and widely recommended algorithm in the practical industry.Most book recommendation systems also use this algorithm.However,the traditional collaborative filtering algorithm,represented by collaborative filtering algorithm,can not deal with the problem of recommendation performance degradation caused by data sparsity.And only using the shallow feature design of the interaction between readers and books,it can not achieve the high-level abstract learning of the relevant attribute features of readers and books.In view of the above problems,this paper uses deep learning technology to build a network recommendation model to solve the traditional recommendation problems and improve the quality of book recommendation.According to the above ideas,the work of this paper is as follows:(1)the traditional recommendation algorithm uses one hot vector to express the data,which can not express the data's own characteristics and fully reflect the data correlation.Moreover,when the amount of data is large,the one hot vector can build the vector matrix,which takes up too much resources.This paper uses the object embedded representation in the data input.Expression can use vectors of low dimension,dense value and real value to represent objects.This method can not only extract the features of the object better,but also reduce the space occupied by itself compared with one hot vector,and is suitable for data input of neural network model.(2)for the information extraction of book title,the traditional recommendation algorithm can not effectively extract the information contained in the text because it is a shallow learning model.In this paper,the text convolution neural network is used to extract the rich information feature expression from the book title.(3)using deep learning to model the probability of readers borrowing books,building a recommendation system model through multi-layer neural network,inputting the features extracted from readers and books into the network,and using multi-layer neural network to deeply interact the features of readers and books,mining the hidden deep interaction between readers and books.(4)build and implement a Book Recommendation Model of deep learning,use the book borrowing data of Nanning university to carry out experiments and compare with the traditional recommendation algorithm,and verify the effectiveness and superiority of the deep learning recommendation model method from the experimental results,so as to add an effective way for the personalized recommendation application of drawing book borrowing in the future.