The ever-changing market demand accelerates the iterative upgrading of new energy vehicles, making understanding user requirements crucial for optimizing product strategies. However, the imbalance in reviews reduces the accuracy of user requirements analysis, which may mislead the purchasing intentions of potential consumers and the product development strategies of manufacturers. To address this, this study develops a BERT-TCBAD-Kano-based requirements analysis method. First, a sentiment analysis method is used to identify user preferences in online reviews. Second, the Text Classification Based on Attribute Dictionary (TCBAD) is proposed to categorize online complaints of users. Finally, user satisfaction and concern are calculated based on preference identification and complaint classification results. User requirements are prioritized based on the idea of the Kano model. Based on the online data of a new energy vehicle, 342 attribute words and 10 user requirements are extracted. The results show that the proposed method improves prediction accuracy by 30 % compared to the traditional Kano model. The method provides significant decision-making support for user-centered product development.