Research On Book Recommendation System Based On Deep Learning

With the development of information technology,the emergence of online book shopping malls has enabled the public to purchase all kinds of book products without leaving home,which greatly promoted the spread of social culture.However,in the face of voluminous books,it is difficult for the public to find books that suit their interests.The traditional book recommendation system has contributed to solving this problem,but the traditional book recommendation system has always suffered from problems such as sparsity and cold start.At the same time,deep learning has received great attention both in industry and in academia.It has achieved great success in areas such as natural language processing,image processing and speech recognition,and is still a great potential.advanced technology.If the deep learning technology can be applied to the book recommendation system to improve the recommendation performance,it will have great development prospects and social significance.In this thesis,the deep learning technology such as multi-layer perceptron and stack denoising auto-encoder is combined with the recommendation system and the auxiliary information feature preprocessing.The main work is as follows:(1)Facing implicit feedback data scenarios,on the basis of in-depth study of neural network-based collaborative filtering algorithm,based on the case that it only uses implicit feedback data and is vulnerable to sparseness problem,an improved book recommendation algorithm based on NCF is proposed.The space is divided into the auxiliary information hidden space and the ID hidden space,and the auxiliary information and the implicit feedback data can be used for learning at the same time.The experimental results show that the proposed algorithm can effectively improve the recommended hit rate and recommended sequence quality,and effectively mitigate the impact of sparsity.(2)Facing explicit feedback data scenarios,after deeply researching the recommendation system based on stacked noise reduction self-encoder,the problem of insufficient recommendation learning ability is limited by using only the stacked noise reduction self-encoder in the previous literature,which restricts the recommendation accuracy.The book recommendation algorithm is improved.The algorithm combines the stacked noise reduction self-encoder and multi-layer perceptron to improve the ability to learn nonlinear interaction functions between hidden vectors.The experimental results show that the algorithm effectively reduces the root mean square error of book recommendation and improves the recommendation accuracy.(3)Finally,this thesis designs and implements a prototype of the personalized book recommendation system based on the two improved book recommendation algorithms proposed in this thesis.The system consists of a user interaction module,a recommendation engine module and a background management module.A complete set of processes from user login to browsing to recommendation is implemented.