| Potential consumers want to find quality decision-aid information from online reviews,car manufacturers want to use online reviews for online marketing,and review sites need to establish a set of evaluation mechanisms and evaluation systems aimed at improving the quality of online reviews.However,the quality of online reviews has not yet formed a scientific and unified evaluation standard,the evaluation indexes and factors influencing the quality of online reviews need to be improved,and the mechanism of the influence of online review quality on consumers’ purchase intention and the mechanism of indirectly promoting product sales is not yet clear.Based on the review of existing literature and theories,this paper summarizes the statistical features of online reviews according to ELM model and statistical analysis methods;extracts the topic features of online reviews by combining the topic mining model and machine learning algorithms;obtains the evaluation indexes of online review quality based on the statistical features and topic features,applies regression analysis to study the effect of online review features on review quality;investigates the contribution of online review quality to consumer purchase intention through structural equation modeling;constructs a car sales prediction model based on machine learning and deep learning algorithms to examine the prediction effectiveness of the model under the influence of online review quality.Through the above studies,the following conclusions are drawn.(1)The analysis of statistical characteristics of online reviews shows that the percentage of reviews in economically developed regions is significantly higher than the percentage of population numbers,and economically developed provinces or provinces with large populations have a higher acceptance of new energy vehicles and a stronger willingness to publish new energy vehicle reviews.In the time dimension,the number of online reviews released and the number of vehicles purchased basically maintain the same growth or fallback trajectory,but the number of reviews released has an obvious lagging cycle.In the frequency dimension,the average time interval between the purchase of a car and the release of a review is 115 days,with the time interval for high-end models being higher than the average and the time interval for mid-and low-end models being lower than the average.Overall,the time interval for reviews of new energy vehicles is more concentrated than that for fuel vehicles.Consumers who buy mid-range new energy vehicles have the most requirements for the purpose of purchasing the vehicle,and those who buy high-end models have the least requirements for the purpose of purchasing the vehicle.(2)The topic features of online reviews show that online reviews of consumers of different model grades include similar topics such as "new energy features","basic features" and "brand features",as well as differentiated topics such as "horizontal comparison" and "individual configuration".The focus of consumers’ online reviews also varies by region.For example,consumers in Northeast,Northwest and North China are particularly concerned about "winter battery performance",while consumers in East China are more concerned about "new energy features" and "intelligence" topics.Consumers in all regions share the same focus on "space" and "appearance".In addition,the high-frequency keywords in online reviews cover all the topic words obtained from topic mining,which also provides a better solution for the quantification of review topics and the evaluation of review quality.(3)The quality evaluation study of online reviews shows that each online review quality evaluation index does not contribute to the quality score to the same extent,and a small number of indicators occupy most of the weight.Review quality shows more of a long-tail distribution in terms of score interval and time span,in which the review quality scores of low-grade models tend to be the most consistent and the least dispersed,while the review quality scores of mid-grade models are the most dispersed and the most discrete.Most online review statistical characteristics have a relatively significant effect on review quality.(4)Research on the sales prediction effectiveness of online review quality shows that review quality has a positive effect on consumer purchase intention.The monthly sales and review quality score monthly trend curves of most models match well.The research results of car sales prediction based on CNN＿LSTM model,LASSO regression model and support vector machine model show that the models are effective in predicting sales of new energy vehicles,among which the CNN＿LSTM model has the highest prediction accuracy.Combining the above research findings,this paper proposes countermeasure suggestions for online review websites,such as enriching car performance scoring dimensions,improving online review publishing rules,and optimizing online review recommendation mechanism;it also proposes countermeasure suggestions for new energy vehicle enterprises,such as strengthening the communication and cooperation of websites,paying attention to the feedback of negative reviews,attaching importance to the processing of high-quality negative reviews,and exploring the value of sales prediction. |
PDF Full Text Request