Research On Recommendation System Based On Emotion Analysis

Along with the updating of computer technology,Internet gradually affects people's life style,compared to the previous offline shopping,now more and more people are going to choice online shopping,this way can really improve the efficiency of shopping,and at the same time,for the consumers,they have already improve their shopping behavior into the consumer experience,as the shopping behavior occurs,it can derive a series of consumer behavior,including product consulting before make the choices,and referring to the historical evaluation of commodities,after shopping,customers will give their reviews and evaluation about the goods,maybe also contains additional comments.With the convenience of the Internet,the variety of commodities in e-commerce platforms is increasing,and the corresponding historical comments on commodities are also increasing,which makes it more difficult for consumers to choose.Meanwhile,it is increasingly important for merchants to quickly obtain consumers' shopping needs and make accurate recommendations.Based on the above problems,this article studies the product recommendation system.Product reviews and product ratings can truly reflect the actual situation of the product,and historical reviews of consumers can truly reflect the personal preferences of consumers.Therefore,through the sentiment analysis of the text data of product reviews,this article proposes a product recommendation system model based on consumers 'emotional preferences.First,using the theme model to extract product attributes.Then,through the sentiment analysis model,the emotional scores of consumers and products on the product attributes are calculated,and the consumer's emotional preferences vector and product attribute sentiment scores vector are obtained.It also introduces the product relevance and consumer relevance,and finally calculates the recommendation degree to judge the products customers are more interested in to realize the recommendation process.The main work of this article is as follows:First,through the analysis of the current status of recommendation systems,this article summarizes the advantages and disadvantages of mainstream recommendation methods,and combined with the current status of sentiment analysis technology,this paper studies the recommendation system based on sentiment analysis and proposes a theoretical model of the recommendation system based on consumer's emotional preference,which is CEPRS.Then,this paper proposes a hierarchical implementation model of the CEPRS.The model implementation is divided into five processes,namely data acquisition,data preprocessing,product attribute extraction,sentiment analysis,and product recommendation.The methods and techniques used in the process are introduced.Secondly,this paper applies CEPRS to book sales recommendation,analyzes the book attributes mined in the existing research,and proposes the significance of subjective attribute extraction of books.Finally,this paper uses the Amazon book review dataset for empirical analysis,mines book attributes in book reviews,constructs a reader preference matrix,and calculates the book's sentiment score matrix based on reader reviews,The CEPRS model is used for recommendation and compared with the recommendation system based on collaborative filtering.The two indicators of recommendation accuracy and recall are used to judge the pros and cons of different models,and the validity of the subjective attributes of the book is verified.