Study Of Reader’s Searching And Borrowing Based On Deep Learning

With the spread of internet, the amount of information and data on Internet is increasing seamlessly. Therefore, people always face the problem of overloaded information on internet, which also exists in book loan system and document retrieval. Searching operation is frequently required when readers want to choose their favorite book and document, however, with the growing quantity, searching consume more and more time, and the searching result can’t always fit the demand of readers, which has a lot of room to improve. Therefore, this thesis is aim to explore the deep learning model to improve the searching technique and book loan system, based on the study of deep learning in machine learning, the current study of book recommendation, as well as recommendation algorithm of clustering and collaborative filtering. Moreover, by the study of deep learning model, the thesis discusses the method and the application in library personalized book recommendation of collaborative filtering based on Restricted Boltzmann Machine(RBM), and collaborative filtering recommendation based on the Deep Boltzmann Machine(DBM). In addition, as for the sort of the result of the document retrieval, the thesis probes literature readers interest degree sorting calculation method using deep belief network(DBN).(1)There are many kinds of personalized recommendation technology at present, the current personalized recommendation technology has been analyzed in this paper. On this basis, collaborative filtering recommendation technology has been studied mainly, and the mixed collaborative filtering method has been discussed based on the clustering of readers and books, the results in the simulation experiments conducted on university students borrow records data sets show that the mean absolute error(MAE) is better than the traditional collaborative filtering recommendation algorithm based on user clustering.(2)Restricted Boltzmann Machine as a deep learning model can be well applied in the collaborative filtering problem. By constructing the reader book interest score matrix, we propose the collaborative filtering book recommend algorithm based on Restricted Boltzmann Machine, and on this basis, the collaborative filtering book recommend algorithm based on Deep Boltzmann Machine is tested and the simulation experiments analysis of the convergence of the two algorithms is made. At last we discuss a method of computing paper interest based on Deep Belief Networks, and test the influence of the Mean Average Precision by the layers of the networks. And the model’s performance of dealing with big scale data has been discussed lastly.