Which approach better samples extreme traffic conflicts? Conventional- vs. machine learning-based sampling methods

Maryam Hasanpour

doi:10.1016/j.aap.2026.108423

Which approach better samples extreme traffic conflicts? Conventional- vs. machine learning-based sampling methods

2026-02-02

0

OA

AI

DOI：10.1016/j.aap.2026.108423

Original

Save

View PDF

Abstract

En 中文

• Extreme conflicts sampled using machine learning methods better reflect conceptual severity levels. • Machine learning methods provide better contextual representation than conventional method. • Extremes classified by isolation forest method more closely preserve characteristics of empirical tail distributions.

Keywords:

Traffic conflicts

Extreme value theory

Sampling techniques

Autoencoder neural network

Isolation forest

Which approach better samples extreme traffic conflicts? Conventional- vs. machine learning-based sampling methods

Abstract

AI Summary

Journal

Organization