Research And Application Of Sentiment Analysis Of New Energy Vehicle Reviews Based On Text Mining

With the rapid development of the Internet today,people can post comments and obtain information through the Internet more conveniently.Therefore,the online reputation of product has become an important factor influencing consumers’ purchasing decisions.With the maturity of technology and the inclination of policies,new energy vehicles have gradually entered the public’s field of vision.The sales volume of new energy vehicles in my country has increased from 18,000 in 2013 to 136.7 in 2020.When buying a car,word-of-mouth reviews will provide consumers with the first impression and influence subsequent car purchase decisions.At the same time,word-of-mouth data serves as the overall evaluation of consumers after the experience of specific models,and auto manufacturers can also understand the market feedback of products based on this,then make improvements and upgrades.Therefore,digging and sorting out the information behind online reviews is of great significance to promote the development of the new energy vehicle market.Firstly,this article first uses web crawler technology and bypasses the anti-crawl mechanism of the website to obtain new energy word-of-mouth comment data on multiple websites such as Autohome,and classifies them according to the purchase price and different models.Then,perform descriptive statistical analysis on the reviews of the new energy vehicles of different price points and different models.Secondly,data cleaning and word segmentation are carried out on the text.In order to be able to segment words more accurately,this paper constructs a dictionary of automobile proprietary names.Draw a word cloud diagram of word-of-mouth comments on new energy vehicles by counting the word frequency of word segmentation.Next,use the word2 vec model to convert the word segmentation result into a distributed word vector,and then use different weighted average word vector methods to obtain the text vector corresponding to each text,and mark the most satisfactory and least satisfactory dimensions in the word-of-mouth comment data as positive emotions and negative emotions respectively.The third step is to put the training data into the machine learning model SVM,LOGISTIC and the deep learning model LSTM to train the marked comment text,and use the test set to verify the model.Through the results of model training,it is found that The comment vector obtained by arithmetically averaging the word vector has a higher accuracy in the classification model than that of the review vector weighted by idf;the classification accuracy of the deep model LSTM is 96.5%,which is higher than that of the machine learning classification model.Then,after using the trained sentiment classification model LSTM to label the unlabeled review data,this article compares the percentages of positive and negative reviews of the various dimensions of new energy vehicle reviews of different price points and models,and obtains that consumers have relatively high negative comments on the interior and comfort of new energy vehicles.Finally,the LDA theme extraction model is used to dig out the focus of consumers’ attention to the interior and comfort and Satisfaction and dissatisfaction dimensions in SUV models.This article applies text mining technology in the field of online reviews of new energy vehicles,and uses online word-of-mouth reviews to mine consumer sentimental evaluations of different dimensions of new energy vehicles.Then,further understand the focus of consumers and the advantages and disadvantages of the products,so as to provide certain improvement suggestions for new energy vehicle manufacturers.Finally,this article points out the possible deficiencies in the research and the issues that need further consideration.