Design And Application Of Movie Recommendation System Based On Variational Matrix Factorization

With the rapid development of the Internet,people can get more and more information,but it also brings the problem of "information overload".People can not get the information which they need in time,and recommendation system is one of the main ways to solve the problem.The recommendation system can effectively help people quickly get the information which they want,but there are some problems such as sparse data and cold start in the traditional recommendation algorithm.In recent years,deep learning technology has achieved good development in image processing and natural language processing.This thesis combines deep learning and the traditional recommendation algorithms,which in order to improve the problems existing in the traditional recommendation algorithm and lift accuracy of the recommendation system.Firstly,this thesis introduces the significance of the recommendation system research,and the principles of several traditional recommendation algorithms and their advantages and disadvantages.This thesis describes several evaluation indicators and similarity calculation methods in the recommendation system.In order to apply the deep learning technology to the traditional recommendation algorithm,this thesis chooses the variational autoencoder(VAE),which is a novel deep network architecture with powerful feature extraction capabilities.By integrating the unsupervised variational autoencoder into the probability matrix factorization,this thesis builds a new recommendation model which is variational autoencoder matrix factorization(VAEMF)through perceptual context.This thesis performs data preprocessing on the description information of the item,and then the VAE captures the context information feature of the processed data,and finally uses the probability matrix factorization to further improve the accuracy of the prediction score.This thesis verifies the model on two public datasets,and experimental results verify the advantages of this method compared with other recommended algorithms,and analyze the important parameters in the model.This thesis uses a technology such as Springboot to build a personalized movie recommendation system,which is a B/S architecture.First of all,this thesis analyzes the requirements of the system,then designs the overall design and specific functions of the system,and then designs the database of the system.At the same time,the development environment used by the system is introduced,and then the modules of the design are realized,and the performance of important modules in the system is optimized to improve the robustness of the system.The system also applies the VAEMF recommendation model proposed in this thesis to the recommended modules of the system.Finally,the main modules of the system are tested and the expected results achieved.