Topic classification of electric vehicle consumer experiences with transformer-based deep learning

Area of Science:

• Environmental science and policy
• Computer science and artificial intelligence
• Transportation engineering

Background:

• The transportation sector significantly contributes to greenhouse gas (GHG) emissions and global health issues.
• Electric vehicles (EVs) are promoted by government policies to reduce GHG emissions.
• Consumer data, often unstructured, is underutilized in EV charging infrastructure planning.

Purpose of the Study:

• To apply advanced deep learning techniques to analyze unstructured EV user review data.
• To identify key topics of consumer attention regarding EV charging infrastructure.
• To enhance public policy decisions and market intelligence for the EV sector.

Main Methods:

• Utilized transformer-based deep learning models.
• Analyzed a nationally representative sample of EV user reviews.
• Achieved high classification accuracy (over 91%) and F1 scores (0.83).

Main Results:

• Successfully extracted topics of attention from unstructured EV user reviews with high accuracy.
• Demonstrated superior performance compared to existing algorithms in the domain.
• Developed a scalable method for public policy analysis in the EV charging market.

Conclusions:

• Deep learning offers a powerful tool for analyzing unstructured consumer data in the EV sector.
• Improved utilization of consumer insights can optimize EV charging infrastructure development.
• This approach provides valuable intelligence for the rapidly expanding EV charging market, projected to reach $27.6 billion by 2027.