Kansei engineering for new energy vehicle exterior design: An internet big data mining approach

Abstract

New energy vehicles (NEVs) such as electronic cars represent a major trend in the automobile industry, where most their exterior designs still follow those of convention fuelled vehicles (FVs). It is important to investigate whether NEV users have unique requirements that differ from those of traditional users. Kansei engineering is a practical tool for perceptual demand analysis. However, the conventional method requires questionnaires or surveys to perform limited data collection. In this study, we utilised massive internet data to collect user Kansei requirements for NEV exterior design. The Scrapy crawler was adopted for data collection and a bidirectional long short-term memory, conditional random field, and multilayer perceptron framework was developed for text mining. To quantify design features and Kansei image scores, a hybrid Apriori + structural equation model (SEM) system is proposed, where the data-driven Apriori algorithm can explore the hidden relationships in big user generated comments, while the SEM model captures the users’ behaviour and decision procedure so that to provide interpretable results. In addition, the association rules mined from user comments by Apriori can facilitate the specification of a complicated SEM model, substantially reducing the modelling and calibration effort. Goodness-of-fit results suggest that the proposed model outperforms conventional models. A case study on 1805 automobiles, 287 brands, and 369105 comments was conducted and the results suggest that some design features that would increase the Kansei image scores for conventional FVs may have the opposite effect on NEVs. Discussions on engineering and managerial insights are presented and the discovered rules and relationships are employed to develop a design-aided system.