Research On Kansei Engineering For The Exterior Design Of New Energy Vehicle Based On User Comment Mining

Within the recent ten years,new energy automobile has obtained more and more attention.However,the design of most new energy automobiles still follows the design of traditional fuel automobiles,particularly the exterior design.It would be of importance to investigate whether consumers who chose new energy automobile have different exterior design requirements from those who chose traditional fuel automobiles.Kansei engineering is widely used to analyze users’ requirement in product exterior design because of its ability to capture and quantify users’ perceptual requirements.The traditional Kansei engineering design method needs to conduct various interviews or experiments,consequently leading to a high cost on data collection.With the development of mobile internet and the culture of opinion sharing,users posted a massive volume of comments on various online communities,which contains rich information about users’ preferences on automobile design.This is a great opportunity for enterprises and designers where they could analyze user needs in depth with higher precision for product design and marketing deployment.To achieve this goal,the paper proposes a research method to employ the Kansei engineering for the new energy automobile’s exterior design.Firstly,we collected user attributes,comments,feedbacks and exterior design parameters by the automated web crawler which was deployed on several automobile websites based on Python programming language and Scrapy automated crawler framework.Then,through the text cleaning,building the corpus of vehicle domain,the text segmentation,training word vector model and other methods,the fast preprocess of data is achieved.Secondly,in order to accurately extract the perceptual needs of users from massive user comments and to improve the credibility of perceptual information,this paper used the deep learning framework Pytorch to build the model for extracting automobile keywords based on the Bidirectional Long Short-Term Memory(BI-LSTM)neural network combined with Conditional Random Field(CRF).This model could extract keywords on automobile appearance(aspect)and user emotional information(opinion).In addition,to obtain the relationship matching between the aspect and opinion,we build a model based on Multi-Layer Perceptron neural network combined with entity information.Thirdly,to help designer to better explore the relationship between product elements,user attributes and user perceptual experience,this paper developed a quantitative analysis method with Kansei engineering,where a hybrid model of structural equation model(SEM)and the data mining Apriori algorithm was proposed.In order to verify the performance of the proposed method,it was compared with Linear Regression model and the conventional SEM method.Results suggested that the proposed method is the best in the sense of goodness-of-fit(R~2).Seventeen fuel vehicles and thirteen electric vehicles were selected as research objects.User comments and feedbacks were collected from various online platforms such as Pacific Automotive Website,Auto Home Website,etc.The collected opinions,user attributes and exterior design parameters were input into the developed Kansei engineering model.Numerous insights were obtained from the model output,suggesting that the proposed method could offer good guidance for product design improvement and marketing research.Finally,we used Python programming language and MySQL database to build a Kansei engineering application system based on the results of our quantitative analysis of Kansei engineering,where the several key interfaces the application were displayed.The developed system could offer a series of suggestions and guidance such as product design requirements for designers,marketing insights for research and purchase recommendation for consumers,showing the potentiality of applications of the proposed method.