Design And Implementation Of Recommendation System Based On Matrix Factorization And Convolution Neural Network

At present,various industries in China are actively embracing the Internet,and ecommerce is playing an increasingly important role in the daily life of the public.However,information overload makes it difficult for users to make purchase decisions.The current main way to solve this problem is to use a recommendation system to recommend personalized products or services to different users based on user and project related information.In the design of the recommendation system,the item 's comment information and the user 's social information are of great significance to the performance of the recommendation system.On the one hand,the comment information is the leading factor for users to make online purchasing decisions in e-commerce,on the other,The user's social information is also very helpful to alleviate the cold start problem of the recommendation system.Although some existing models use review information to improve the performance of the recommendation system,and at the same time consider the impact of the credibility of the review information on the recommendation quality,a neural attention regression model(Neural Attentional Rating Regression with Reviewlevel Explanations,NARRE)with an evaluation level explanation is proposed,But the model does not consider the user's social information,and its recommendation performance still has room for improvement.Therefore,the main work of this paper is to improve the original NARRE model,add the user's social information to the data source,and then extract three different features with the neural network model: first,extract the user's preference features from the user's comment information;second,extract the implicit features from the project's comment information;third,extract the hidden features from the user's social network Take the social characteristics of users.Finally,the above three features are combined with the Matrix factorization model to get the final recommendation results.The whole model can alleviate the sparse problem and cold start problem of the algorithm by introducing user social information,so as to improve the performance of the recommendation algorithm.Specifically,the main work of this paper is as follows:(1)A multi-source heterogeneous data recommendation model combining Matrix factorization and neural network is proposed.Based on the existing NARRE model,the user's social information is added to the data source,so that the model can provide the same comment level explanation as the NARRE model(scoring the comment usefulness,selecting the high useful comments),and the recommendation performance is further improved.(2)In the design of the prediction layer,the traditional Matrix factorization technology is combined,and the user preference feature,project feature and user social feature extracted by the neural network model are integrated with the Matrix factorization model,which alleviates the sparse problem and cold start problem of the recommendation model,so as to improve the performance of the recommendation algorithm.(3)According to the proposed multi-source heterogeneous data recommendation model combining Matrix factorization and neural network,a restaurant recommendation system is designed and implemented.The data set uses yelp data set to realize the functions of user management,restaurant search,restaurant information viewing and user personalized recommendation.Through the experiment of yelp benchmark data set in the restaurant field,it shows...