The Research And Implementation Of Rating Prediction Method Based On Text Sentiment

With the rapid development of Internet and E-commerce,online shopping has gradually replaced traditional shopping and become the main way of shopping.How to effectively use the data of user's comments on goods or services to discover user's interests and preferences and improve the accuracy of product recommendation has become a hot research topic.This paper focuses on sentiment analysis and prediction based on user reviews.Through the analysis to determine the user's emotional polarity and intensity of different characteristics of the product,and establish a prediction model to predict the user's rating of a comment information.In addition,the paper also analyzes the commodity characteristics that affect the polarity of user reviews,which can help producers make decisions and recommend related products to specific users.Based on text sentiment analysis,this paper establishes XGBoost integrated learning prediction model by analyzing the feature words and emotion intensity of commodities,a prototype system of user rating prediction based on product reviews is designed and implemented.The main work of this paper is as follows:1.This paper analyzes the background significance and research status of commodity rating prediction.This paper studies the algorithm and model of feature words and opinion words extraction and prediction,and discusses the feasibility of commodity rating prediction.2.The improved HITS algorithm is used to extract feature words and opinion words,and the method of extracting feature words and opinion words separately is improved.The mutual information is used as the edge weight of the improved HITS algorithm,and the feature words and opinion words are extracted by pairing.The numerical results show that the improved HITS algorithm can effectively improve the accuracy of feature words and opinion words extraction.3.A quantization method of sentiment intensity of opinion words is proposed.The core of the algorithm is to count the probability distribution of opinion words in different ratings,and use the mean value as the final emotional strength.In this method,the intensity values of words with different features but with the same opinion can be transformed,and the emotional intensity and polarity can be expressed at the same time4.The XGBoost integrated prediction model is trained by taking the commodity feature as the training feature and the quantized emotion intensity as the value of the feature.The accuracy and efficiency of the prediction model are analyzed by experiments.5.Design the implementation of the prediction system,including the system architecture,the implementation of different modules,the design of the database and the display of the visual interface.