Driving behavior is one of the most critical factors in traffic accidents. Accurate vehicle acceleration prediction approaches can promote the development of Advanced Driving Assistance Systems (ADAS) and improve traffic safety. However, few prediction models consider the characteristics of individual drivers, which may overlook the potential heterogeneity of driving behavior. In this study, a vehicle acceleration prediction model based on machine learning methods and driving behavior analysis is proposed. First, the driving behavior data are preprocessed, and the relative distance, relative speed, and acceleration of the subject vehicle are selected as feature variables to describe the driving behavior. Then, a finite Mixture of Hidden Markov Model (MHMM) is used to divide the driving behavior semantics. The model can divide heterogeneous data into different behavioral semantic fragments within different time lengths. Next, the similarity of different behavioral semantic fragments is evaluated using the Kolmogorov–Smirnov test. In total, 10 homogenous drivers are classified as the first group, and the remaining 20 drivers are classified as the second group. Long Short-Term Memory (LSTM) and Gate Recurrent Unit (GRU) are used to predict the vehicle acceleration for both groups. The prediction results show that the proposed method in this study can significantly improve the prediction accuracy of vehicle acceleration.

The empirical Bayes (EB) method is commonly used by transportation safety analysts for conducting different types of safety analyses, such as before–after studies and hotspot analyses. To date, most implementations of the EB method have been applied using a negative binomial (NB) model, as it can easily accommodate the overdispersion commonly observed in crash data. Recent studies have shown that a generalized finite mixture of NB models with K mixture components (GFMNB-K) can also be used to model crash data subjected to overdispersion and generally offers better statistical performance than the traditional NB model. So far, nobody has developed how the EB method could be used with finite mixtures of NB models. The main objective of this study is therefore to use a GFMNB-K model in the calculation of EB estimates. Specifically, GFMNB-K models with varying weight parameters are developed to analyze crash data from Indiana and Texas. The main finding shows that the rankings produced by the NB and GFMNB-2 models for hotspot identification are often quite different, and this was especially noticeable with the Texas dataset. Finally, a simulation study designed to examine which model formulation can better identify the hotspot is recommended as our future research.

Wildlife‒vehicle collision (WVC) data usually contain two types: the reported WVC data and carcass removal data. Previous studies often found a discrepancy between the number of reported WVC and carcass removal data, and the quality of both datasets is affected by underreporting. Underreporting means the number of WVCs is not fully recorded in the database; neglecting the underreporting in WVC data may result in biased parameter estimation results. In this study, a copula regression model linking wildlife‒vehicle collisions and the underreporting outcome was proposed to consider the underreporting in WVC data. The WVC data collected from 10 highways in Washington State were analyzed using the copula regression model and the Negative Binomial (NB) model. The main findings from this study are as follows: (1) the Gaussian copula model can provide different modeling results when compared with the conventional modeling approach; (2) the hotspot identification results indicate that the Gaussian copula-based Empirical Bayes (EB) method can more accurately identify hotspots than the NB-based EB method. Thus, the proposed copula model may be a better alternative to the conventional NB model for modeling underreported WVC data.

The time it takes to clear road traffic incidents is an important performance measure in highway management. Understanding the effects of explanatory variables on incident clearance time is a necessary step in developing incident management strategies. Previously, hazard-based duration models have been used extensively in identifying the influential factors in incident duration. Quantile regression models have been introduced in survival analysis, and these offer a more flexible approach to analyse incident duration data. The advantage offered by quantile regression lies in the ability to make inferences about the effect of explanatory variables on different quantiles of the incident duration distribution. In this work, quantile regression was applied to an incident clearance time dataset collected on freeway sections in Seattle, USA. Based on the estimation results from quantile regression, four influential factors were identified to have a stronger impact at the upper tail of the clearance time distribution. These factors cannot be identified through the conventional hazard-based duration methods used in most previous studies. The findings of this study suggest that the additional information obtained from quantile regression about the short-term and long-term effects of explanatory variables is potentially useful for assigning clearance priority to different road traffic incident types.