Car Reviews Sentiment Analysis Based On Text Mining

As of March 2021,the domestic car ownership is close to 300 million,and the car is gradually transforming from the original asset goods to consumer goods.For different types of consumers,their requirements for new car of space,power,control and others are different,and they always selecte a few favorites to compare in the price range.For automakers,learning about consumer preferences from website big data can focus on some of the car’s facilities in the future and release models that meet more market demand.Therefore,emotional analysis of automobile reviews can capture key indicators from massive data,generate word cloud maps for automobile reviews with different price ranges,and reflect positive and negative emotional information.It can intuitively understand the real needs of consumers and help automobile manufacturers to make correct decisions.A total of 32,648 comments were collected from the word-of-mouth module of Automobile House website,and then were processed,including removing duplicate texts,Jieba word segmentation,deactivating words,retaining Chinese and English texts.Then we use TF-IDF、Word2vec、Gini index and Chi-square test for feature extraction.Comparing the four feature extraction methods,we find that the performance of Word2 vec on the four classifiers is optimal,which shows that the use of Word2 vec to extract feature words improves the accuracy of text emotion classification to a certain extent.Using support vector machine,KNN-based Bagging method,Ada Boost,naive Bayes as base classifier,A Stacking integration model based on logical regression is0.968.And the values of emotion analysis by other models on the sample dataset are arranged from large to small,respectively,support vector machine,Ada Boost 、 K nearest neighbor algorithm,naive bayes.And then in the price range(0,10],(10,20],(20,35],(35,50],(50,+∞),a Stacking integration model based on logical regression.Classification and prediction of comments on mainstream car and SUV are carried out,and objective evaluation of users is identified according to the Word Cloud generated by positive and negative emotions.For consumers to buy cars and car enterprises to further expand the market it provided a certain reference.