Modeling Short‐Term Highway Traffic Conflict Frequency: Integrating Interaction Effects With Stochastic Components

The application of traffic conflicts in road safety assessment is increasingly favored, owing to its critical importance in conducting real‐time safety analyses and formulating proactive safety management strategies. The study aims to construct an accurate model to predict the frequency and occurrence of traffic conflicts on highways within a short time frame. Utilizing the highD dataset, we construct traffic characteristic indicators as covariates, including traffic volume (TV), speed variation coefficient (SVC), the proportion of large vehicles (PLV), lane‐to‐lane average speed difference (LASD), and front‐to‐rear vehicle average speed difference (FASD). Traffic conflict frequency, measured by the time‐to‐collision threshold of fewer than 4 s, serves as the dependent variable. The study employs zero‐inflated Poisson, zero‐inflated negative binomial, hurdle Poisson, and hurdle negative binomial models to fit conflict frequency, accounting for interaction effects, random effects, and random parameters. The findings indicate that models incorporating interaction effects yield a better fit than those not considering interaction effects. Furthermore, models considering interaction effects, random effects, and random parameters show superior fit compared to models only considering interaction effects, albeit with increased complexity under the hurdle Poisson and hurdle negative binomial distributions. These results offer valuable insights for managing and controlling highway traffic safety.