Research And Implementation Of Score Prediction Algorithm Based On Transformer

With the development of Internet technology,people are faced with explosive growth of network data and information,which is often difficult to receive and process effectively.In order to make more targeted use of these data,recommendation system comes into being,which brings convenience for users to obtain personalized information.In the field of recommendation system,rating prediction is one of the core research directions.Past collaborative filtering algorithms,such as problems such as cold start,although its and deep learning technology of fusion of the above problems to some extent,but there are still inadequate feature extraction of users and commodities,not good two-way interactive user and commodity information,such as limit,to some extent,affected the score recommended effect on the accuracy of the prediction algorithm and interpretability.Aiming at the problem of rating prediction in recommendation system,this paper completes the following three parts:1.Design and implement a rating prediction algorithm DAM4R based on Transformer.This paper,based on Transformer’s model structure,processes user and product review information bidirectional and relates it to product codes to predict product ratings.The algorithm in this paper maximizes the extraction and utilization of the two types of comment features to improve the interpretability of the scoring prediction algorithm.2.A score prediction algorithm IS4R based on Transformer is proposed.In order to further improve the accuracy of grading prediction algorithm,the comment text to handle user and goods this specific problem,adjust the coding structure of the Transformer,the comment and commodity information extracted from users with related deep vector,make full use of the information users and the number of goods and score,the predicted ratings at the same time improve the accuracy of grading.3.Based on the designed scoring prediction algorithm,this paper uses open data set to recommend products for users and realizes a product recommendation system that can demonstrate the effect of scoring prediction algorithm.The system can meet the basic requirements of the recommendation system,can be integrated into the scoring prediction algorithm,the user to make the corresponding product prediction,intuitive display of the recommendation results.Through comparative experiments on Yelp and Amazon data sets,it is verified that DAM4R algorithm proposed in this paper improves the interpretability of rating prediction algorithm.At the same time,the accuracy of the algorithm is compared on Yelp and Amazon data sets,which verifies that the IS4R algorithm proposed in this study improves the accuracy of the score prediction results to a certain extent.The product recommendation system can meet the basic needs of users and realize the visual display of the recommendation effect of the scoring prediction algorithm.