Systematic Framework for Classification and Prediction of Post-Crash Delay in Freeways

Abstract

This study develops a systematic framework for analyzing and predicting post-crash delays on freeways using data from Interstate 4 in Florida, U.S. Through a Bayesian Gaussian mixture model, the study classified post-crash delays based on congestion parameters, travel time reliability metrics, and traffic flow characteristics across multiple upstream segments. Logistic regression analysis revealed several key factors, such as pre-crash traffic conditions. In particular, higher volume and greater speed variation significantly increased the likelihood of severe post-crash delay. In addition, crashes during off-peak hours, fatal crashes, and those involving commercial vehicles increased the probability of severe post-crash delays. The study developed delay prediction models for three consecutive 15 min intervals following crashes, achieving good R² values. For the initial 15 min interval, pre-crash delay and travel time variability were the strongest predictors. The 16–30 min interval predictions were heavily influenced by first-interval traffic patterns, while the 31–45 min interval predictions relied primarily on second-interval conditions. The findings play a critical role in advancing proactive integrated corridor management strategies by enabling traffic management authorities to anticipate post-crash delay patterns and implement prompt measures, such as traffic diversions, particularly when pre-crash conditions indicate the risk of severe delay.