Research On Optimization Of Personalized Product Recommendation Algorithm Based On User Reviews

The continuous development of information technology and mobile network makes the online shopping industry thrive.The problem of "information overload" is becoming more and more serious due to the huge trading volume and the increasing variety and quantity of products.Moreover,related research shows that the Internet has a profound impact on consumer behavior patterns,and more and more consumers are accustomed to searching for relevant product introductions and other user reviews on the Internet before making purchasing decisions.The personalized recommendation system is seen as an effective way to solve the above problems.It recommends products and services to target users based on their interest characteristics and purchase behaviors.In this field,the mining based on user ratings,such as the traditional collaborative filtering algorithm,has been studied and applied most widely.However,with the surge of user comment data,relevant scholars found that simple ratings could not accurately represent user preferences,so they considered introducing comment data to assist analysis.Compared with foreign countries,there is still more room for the development of Chinese comment corpus mining.Considering the importance of comment corpus to personalized recommendation system,this thesis,on the basis of the existing opinion mining research,focuses on the topic mining of user comments,introduces emotional factors,builds a topic and sentiment hybrid model,and combines user ratings to generate recommendations,and proposes a personalized product recommendation algorithm based on user reviews.First of all,this study systematically summarizes and sorts out a large number of documents in related fields.This thesis summarizes the related concepts,common algorithms and evaluation indicators of personalized recommendation system,and studies the concept,method and fine-grained comment mining method of comment mining,which lays a theoretical foundation for the subsequent research.Secondly,this thesis considers introducing emotional factors on the basis of summarizing the user review topic mining research,constructs a four-layer structure of "document-emotion-topic-word" based on the LDA theme model,then analyzes user comment data from two dimensions of emotion and theme.Thirdly,,this thesis combines with the excavated emotion-theme distribution based on the user's rating data,so the traditional collaborative filtering algorithm is improved,and a personalized product recommendation algorithm based on user reviews is proposed.Finally,this study crawls a large number of mobile phone ratings and comment data on the Jingdong self-operated platform,selects the experimental index design comparison experiment,and predicts the accuracy and user satisfaction of the personalized product recommendation algorithm proposed in this thesis.