Research On Data Mining Of E-commerce Reviews Based On Differential Evolution Particle Swarm Optimization

With the progress and rapid development of Internet technology in China,online shopping has gradually become the most mainstream way of shopping,which has brought great changes to people's daily life and work in modern society,and also makes people's daily life and work more convenient and fast.However,the variety of online products is daxxling,and consumers can't feel and experience the products directly,which makes it difficult for them to make purchase decisions.The user comment information in e-commerce platform is the subjective feeling of the consumers who have purchased the goods after personal experience,which can provide certain reference for potential users.However,the e-commerce user group is becoming larger and larger,the amount of product comment information is increasing rapidly,and the display strategy of platform comment information is unreasonable and other reasons make it difficult for users to extract valuable information from it.The main contents of this paper are as followsFirst,data acquisition and preprocessing.This paper first describes the current commonly used crawler technology,through the use of web crawler technology to collect the review text data of cultural and creative products in the Palace Museum of e-commerce platform,and through a series of technologies such as data cleaning,subjective sentence extraction,Chinese word segmentation,removal of stop words and so on,carries out the relevant preprocessing operation on the review text,improves the effectiveness of the data,and makes a contribution to the following clustering analysis work We are fully prepared.Second,Word2Vec model is used to obtain word vector.Short text data contains few words,which makes feature extraction difficult.Using traditional text clustering model for short text clustering analysis,often can not get the ideal cluster structure,which is not conducive to the follow-up application research.Word2Vec word vector model can transform the central word item into a word vector in a word space through the context content of the central word item.Compared with the traditional vector space model,it considers the influence of semantic environment on the training results,and has certain advantages.The idea of Word2Vec word vector model is that words with similar context should have similar semantics,so that the corresponding word vectors in the word space with similar semantics are closer.Thirdly,using the clustering results of feature words and emotional tags to construct the dimensions of commodity features and emotions,the characteristics of the commodities concerned by consumers and their subjective feelings are mined from the two perspectives of features and emotions.The dimension system provides an objective and effective reference standard for users to make purchase decisions,and the feasibility and effectiveness of this method are verified by experiments.The experimental results show that the improved algorithm based on the similarity matrix of clustering center vectors between individuals can improve the accuracy of short text data processing,and it has more advantages than the traditional clustering algorithm.Compared with the traditional algorithm,the clustering effect of the improved algorithm is better,and the convergence speed of the algorithm is faster.