Research On Catering Personalization Analysis And Recommendation Model Based On Neural Network

With the popularization of the Internet and the widespread application of the mobile Internet,online catering platforms represented by Dianping and Meituan have developed rapidly,and online catering platforms have accumulated massive amounts of user comment data.Analyzing users’ comment through reasonable and efficient technical means,obtaining user personalized information,and providing users with reliable recommendation results according to user personalized needs,has important commercial value.This paper studies the user personalized analysis and recommendation model of online catering platforms.The research is include two parts:(1)User personalized analysis.In terms of user personalized analysis,this article mainly adopts the aspect-level sentiment analysis method,and uses the user comment personalized analysis neural network model based on multi-task learning and hierarchical attention mechanism.This model can give full play to the advantages of neural network technology to process massive user comment text information.Through aspect-level sentiment analysis technology,we can dig out users’ attitudes towards multiple aspects of restaurant services from user review texts,help businesses better understand user needs,and improve production and business activities.(2)Personalized recommendation model.In terms of recommendation model,this paper designs a collaborative recommendation model that integrates aspect-level sentiment analysis,and converts the result of aspect-level sentiment analysis into a multi-attribute score.The principle of information entropy is used to measure user multi-attribute rating features and restaurant multi-attribute rating features,and the multi-attribute rating features are spliced with the user rating vector and restaurant rating vector respectively,and the deep semi-automatic edge encoder is used for learning to obtain the hidden features in the middle.And use the matrix factorization model to get the recommendation result.