Research On The Application Of Online Comments Based On Text Mining

The rapid development of Internet e-commerce has made online shopping a trend.Online shopping has brought opportunities and challenges to major e-commerce platforms and manufacturers.In addition to developing and upgrading products to drive consumption,merchants also need to understand the user 's voice that can meet their real needs as much as possible.With the rise of online shopping,consumers have used the online shopping platform to publish their shopping experience and opinions on goods and services has become a trend,and a large amount of comment data has also been generated,and these review texts contain a lot of valuable potential information,so through the analysis of product review information,on the one hand,it can provide a strong basis for businesses to find product shortcomings,improve product quality,and improve service attitudes.On the other hand can enable consumers to fully understand the product is conducive to purchase.The research object of this article is the online review data of laptops.First,the review data of Huawei Honor MagicBook2019 and ASUS Vivobook ultra-thin laptops are collected from Jingdong Mall using the Octopus Data Collector for analysis.Next,the text data is cleaned and pre-processed,including text deduplication,chinese word segmentation,and stop words,etc.,and then the processed text is vectorized and expressed as structured data that can be recognized by the computer.In the process of researching the sentiment tendency in this paper,considering that supervised learning methods need to have already annotated text,this paper uses a combination of dictionary construction and machine learning,which not only solves the cumbersome problem of manual annotation,but also adds three text classification algorithms,such as support vector machines,K-nearest neighbor and Naive Bayes.Firstly establish a classification model on the training set and then apply the model to the test set,and evaluate the classifier through three indicators: precision rate,recall rate,and F₁ value to obtain the optimal The classifier model of is a support vector machine;at the same time,this article also analyzes the characteristics of consumer review data,uses word cloud visualization technology to have a preliminary understanding of the characteristics of the product,and then positive and negative reviews of the two laptops Establish LDA theme models separately,using the cosine distance between topic vectors,the optimal number of topics can be obtained by R programming.On this basis,summarize the consumer's views on the product,and find that consumers are mainly concerned about the speed,appearance,portability,system,performance,screen display effect,and after-sales service of the computer.Finally,combined with the difference in the results of text mining analysis,it provides targeted suggestions for the manufacturers of the two brands,and at the same time provides a certain reference for consumers to make purchase decisions based on the differences between the advantages and disadvantages of the computer.