Research on Online Review of Commodities
Download as PDF
DOI: 10.23977/GEFHR2020.007
Author(s)
Xintian Caia, Wen Wen
Corresponding Author
Xintian Caia
ABSTRACT
First, we performed data preprocessing including cleaning and filtering the data, performing independent analysis on the data, etc., removing redundant tags, extracting high-frequency words, classifying them using LDA natural language classification model, and finally identifying 7 topics. In order to analyze the relationship between the review text and the rating, we built a review model and qualitatively and quantitatively analyzed the characteristics of the text review. We analyze it through lexical network analysis, a topic model based on word2vec, and sentiment classification of unsupervised reviews. Then we built a comprehensive product scoring model to score products. The product comprehensive analysis model combines the analytic hierarchy process (AHP) and entropy weight method to improve accuracy; the model also introduces sales volume adjustment parameters to measure the sales popularity of a product among similar products.
KEYWORDS
Data mining, SPASS, comment text