Decoding dissatisfaction drivers and sentiment trends in BEV demand: An integrated LDA and deep learning analysis

This study investigates user dissatisfaction factors and sentiment trends related to battery electric vehicles (BEVs), aiming to uncover nuanced consumption characteristics. While prior research has primarily emphasized attributes such as driving range, charging infrastructure, and price, this paper delves deeper into the factors contributing to user dissatisfaction. Data were collected from user-generated online reviews on Autohome, offering valuable insights into BEV user experiences. The Latent Dirichlet Allocation (LDA) topic model was employed to identify the attributes prioritized by users, followed by the application of ALBERT to achieve a refined classification of these attributes. Additionally, a Temporal Convolutional Network-Bidirectional Long Short-Term Memory (TCN-BiLSTM) model was utilized to predict sentiment dynamics over time. The findings provide a theoretical framework for industry stakeholders to develop a comprehensive understanding of user behavior. This research contributes actionable insights for formulating strategies aimed at fostering the stable growth and efficient operation of the market and industry.