Sentiment Analysis Of Online Comments On New Energy Vehicles Based On Text Mining

With the continuous improvement of technology and policy support,the market share of new energy vehicles has been expanding,and sales have been increasing year by year.As of September 2022,the sales of new energy vehicles in China had exceeded 4million units,with a penetration rate of over 25%.At the same time,the popularity of the Internet has led to an exponential growth of network information,making it more convenient for people to express their opinions and obtain information through the network.During the car purchasing process,online comments provide consumers with product information and influence their purchasing decisions.In addition,consumer comment can provide objective evaluations of user experiences for automobile manufacturers,help them understand product feedback in the market,and make targeted improvements to enhance their competitiveness,promote the vigorous development of the new energy vehicle industry.This paper first divides new energy vehicles into four price levels based on price and crawls the top ten car models’ comment data on each price level on the autohome website in the second half of 2022.Secondly,the comment texts are cleaned and segmented,and professional automobile dictionaries and stop-word dictionaries are introduced to improve segmentation accuracy.The word2vec model is then used to train the word vector of the segmented result to obtain a text vector corresponding to each text.Then,the most satisfied and dissatisfied dimensions of the comment texts are respectively used as positive and negative emotional texts.They are divided into training and testing sets at an 8:2 ratio,and the data is trained and verified using Logistic regression,SVM,random forest,and Adaboost models.After comparing the results of each model,the SVM model performs the best,with an emotional classification accuracy of 94%.Finally,the trained SVM emotional classifier is used to predict the emotional classification of untagged comments and compared and analyzed the prediction results.Based on the preliminary analysis results,positive and negative emotional comments on cost-effectiveness and driving experience,negative emotional comments on the interior and comfort,and positive and negative emotional comments on the four price-level car models are selected for LDA topic extraction analysis to dig deeper into consumer concerns and product pros and cons in different dimensions.Through the study,this paper draws the following conclusions: First,the reputation feedback of new energy vehicles in terms of cost-effectiveness,driving experience,and appearance is good.Second,most consumers are dissatisfied with the interior and comfort of new energy vehicles.Third,there are significant differences in consumers’ evaluation focus and attention for new energy vehicles of different price levels.Fourth,there is still significant room for improvement in the endurance,energy consumption,and intelligence of new energy vehicles.The innovation of this paper lies in the transformation of traditional data collection and processing methods into the use of portal website network comment information and the use of natural language processing technology for implicit value information extraction.At the same time,the emotional analysis and topic extraction of new energy vehicles in this paper are not only limited to the analysis of overall evaluations but also divided into multiple price levels for analysis,proposing more targeted improvement strategies.