| With the development of the Internet and e-commerce and the rise of online shopping,e-commerce platforms have accumulated a large amount of review texts.The recommendation system is the core of the e-commerce platform.With the accumulation of products,the e-commerce platform has a large number of products of the same category,but in fact,the characteristics and characteristics of each product are different.It should be based on the user’s preference and preference for product features.In this context,this paper proposes an improved collaborative filtering algorithm based on the similarity of user preferences based on e-commerce review text.The method in this paper firstly extracts sentiment scores from review text combined with product feature extraction based on the LDA topic model and sentiment analysis based on sentiment dictionary.Then based on the heuristic similarity,the user preference similarity is proposed,and then the joint similarity is constructed.Combining the comment text information to improve the similarity calculation of the collaborative filtering algorithm,it can improve the traditional collaborative filtering algorithm that only calculates the similarity through the user-product rating matrix,but lacks the shortcomings of identifying the user’s specific preferences.It is found that the preferences at the product feature level are more similar Neighbors of users.The data in this article comes from a large English e-commerce review data set Amazon.In this paper,benchmark models for comparison are user-based collaborative filtering algorithm(ACOS-User-CF),and PIP-based CF(PIP-based CF),NHSM-based collaborative filtering algorithm(NHSM-based CF)。The experimental results show that compared with the K-nearest neighbor collaborative filtering algorithm based on traditional similarity,the improved user preference similarity method based on review text information proposed in this paper improves the accuracy of recommendation. |
PDF Full Text Request