A science-based approach to classifying light vehicles in Europe: methodology and case studies

Area of Science:

• Transportation Science
• Statistical Modeling
• Vehicle Engineering

Background:

• Current European light vehicle classification relies on outdated empirical methods, failing to capture fleet evolution.
• Existing methods lack transparency and do not adequately address environmental, safety, and urban planning impacts.
• Accurate vehicle categorization is essential for understanding road transport sector dynamics.

Purpose of the Study:

• To develop a scientific and reproducible method for classifying light vehicles using a Bayesian statistical approach.
• To establish transparent criteria for vehicle segmentation, prioritizing explainability over complex machine learning models.
• To provide a robust framework for updating vehicle fleet models and supporting multi-purpose classification.

Main Methods:

• Utilized a Bayesian statistical method for explicit and reproducible light vehicle segmentation.
• Identified key physical vehicle attributes to define clear boundaries between segments.
• Employed linear relationships between variables to establish interpretable classification criteria.

Main Results:

• The proposed Bayesian algorithm achieved 82% accuracy in assigning vehicles to original segments.
• Demonstrated comparable accuracy to unsupervised machine learning models while offering superior transparency.
• Successfully revealed clear, interpretable boundaries between different vehicle segments.

Conclusions:

• The Bayesian approach offers a scientifically sound and transparent alternative for light vehicle classification.
• Findings support the updating of vehicle fleet models, especially for environmental and energy consumption analyses.
• The method provides a potential standard for versatile vehicle classification in transport research and policy.