Ensuring traffic safety in transportation is paramount, as road accidents profoundly impact public well-being and a region’s economic stability. Among various aspects of traffic safety, crash severity analysis is a vital concern for researchers and transportation agencies. This project focuses on predicting crash severity through machine learning, utilizing a six-year dataset from Southeast Michigan. The goal is to develop models and identify influential variables, offering insights for researchers and transportation agencies. Employing a two-step methodology, integrating correlation analysis and machine learning models, various classifiers, including Decision Tree (DT) and Random Forest (RF), were applied. Results show DT’s computational efficiency superiority, with DT and RF outperforming other models. However, aggressive up-and-down sampling for data imbalance led to poor accuracy. This research significantly contributes to transportation safety with broader implications for public safety and policy development.
Traffic Safety through Machine Learning: A Study of Crash Severity Factors
2024 IEEE World AI IoT Congress (AIIoT) ; 0016-0023
2024-05-29
998752 byte
Conference paper
Electronic Resource
English
Discovering Crash Severity Factors of Grade Crossing With a Machine Learning Approach
British Library Conference Proceedings | 2019
|Taylor & Francis Verlag | 2021
|Examining traffic crash injury severity at unsignalized intersections
Online Contents | 2010
|