A systematic review of statistical models and outcomes of predicting fatal and serious injury crashes from driver crash and offense history data

Identification and screening of key traffic violations: based on the perspective of expressing driver's accident risk

Drawing on the core idea of Propensity Score Matching, this study proposes a new concept named Historical Traffic Violation Propensity to describe the driver's historical traffic violations, and combines the new concept with an improved mutual information-based feature selection algorithm to construct a method for screening key traffic violations from the perspective of expressing driver's accident risk. The validation analysis based on the real data collected in Shenzhen demonstrated that drivers' state of Historical Traffic Violation Propensity on 19 key traffic violations screened have a stronger predictive ability of their subsequent accidents compared to the level in existing research. The positive state of Historical Traffic Violation Propensity on 'Drinking', 'Parking in dangerous areas', 'Wrong use of turn lights', 'Violating prohibited and restricted traffic regulations', and 'Disobeying prohibition sign' will increase the probability of a driver's subsequent accident by more than 1.7 times. The research provides directions to more efficiently and accurately capture the driver's accident risk through historical traffic violations, which is valuable for identifying high-risk drivers as well as the key psychological or physical risk factors that manifest in daily driving activities and lead to subsequent accidents.

Looking back in the rearview: Insights into Queensland's rear-end crashes

OBJECTIVE

Rear-end crashes cost the most out of any crash type to the Compulsory Third Party insurance scheme of Queensland. Rear-end crashes are also one of the most common types of crashes, peaking as the most common type of crash in Queensland in 2019. This study investigated the environmental, driver, and vehicle characteristics associated with rear-end collisions in Queensland, Australia.

METHODS

A totle of 367,230 crashes in the Queensland crash database between 2001 and 2021 were used to identify crash trends in Queensland, particularly related to rear-end crashes. For crashes between 2015 and 2021, additional information was gathered and allowed further analysis of factors contributing to rear-end crashes using the quasi-induced exposure method. Two binary classification tree models were used. Model 1 investigated the environmental characteristics that predicted a crash being a rear-end or not. Model 2 investigated the controller and vehicle characteristics that predict a rear-end crash-involved vehicle to be either the striking (at-fault) or struck (not-at-fault) vehicle.

RESULTS

Model 1 indicated that rear-end crashes were overrepresented in urban areas and major cities and where speed limits exceeded 50 km/h. Model 2 indicated that occupants of struck vehicles (not at-fault) were more likely to be severely injured than the striking (at-fault) vehicles in rear-end crashes. License type and vehicle type also influenced the frequency and severity of rear-end crashes.

CONCLUSION

Rear-end crashes can be unjust in that the not-at-fault party is typically more severely injured. Results from the current study suggest that keeping speed limits no higher than 50 km/h could help reduce the severity of rear-end crashes. Increased enforcement of safe following distances could also assist in preventing road trauma as a result of rear-end crashes.

A Crash Data Analysis through a Comparative Application of Regression and Neural Network Models

One way to reduce road crashes is to determine the main influential factors among a long list that are attributable to driver behavior, environmental conditions, vehicle features, road type, and traffic signs. Hence, selecting the best modelling tool for extracting the relations between crash factors and their outcomes is a crucial task. To analyze the road crash data of Milan City, Italy, gathered between 2014–2017, this study used artificial neural networks (ANNs), generalized linear mixed-effects (GLME), multinomial regression (MNR), and general nonlinear regression (NLM), as the modelling tools. The data set contained 35,182 records of road crashes with injuries or fatalities. The findings showed that unbalanced and incomplete data sets had an impact on outcome performance, and data treatment methods could help overcome this problem. Age and gender were the most influential recurrent factors in crashes. Additionally, ANNs demonstrated a superior capability to approximate complicated relationships between an input and output better than the other regression models. However, they cannot provide an analytical formulation, but can be used as a baseline for other regression models. Due to this, GLME and MNR were utilized to gather information regarding the analytical framework of the model, that aimed to construct a particular NLM.

Modelling the relationship of driver license and offense history with fatal and serious injury (FSI) crash involvement

INTRODUCTION

Previous research has indicated that increases in traffic offenses are linked to increased crash involvement rates, making reductions in offending an appropriate measure for evaluating road safety interventions in the short-term. However, the extent to which traffic offending predicts fatal and serious injury (FSI) crash involvement risk is not well established, prompting this new Victorian (Australia) study.

METHOD

A preliminary cluster analysis was performed to describe the offense data and assess FSI crash involvement risk for each cluster. While controlling demographic and licensing variables, the key traffic offenses that predict future FSI crash involvement were then identified. The large sample size allowed the use of machine learning methods such as random forests, gradient boosting, and Least Absolute Shrinkage and Selection Operator (LASSO) regression. This was done for the 'all driver' sample and five sometimes overlapping groups of drivers; the young, the elderly, and those with a motorcycle license, a heavy vehicle license endorsement and/or a history of license bans.

RESULTS

With the exception of the group of drivers who had a history of bans, offense history significantly improved the accuracy of models predicting future FSI crash involvement using demographic and licensing data, suggesting that traffic offenses may be an important factor to consider when analyzing FSI crash involvement risk and the effects of road safety countermeasures.

CONCLUSIONS

The results are helpful for identifying driver groups to target with further road safety countermeasures, and for showing that machine learning methods have an important role to play in research of this nature. Practical Application: This research indicates with whom road safety interventions should particularly be applied. Changes to driver demerit policies to better target offenses related to FSI crash involvement and repeat traffic offenders, who are at greater risk of FSI crash involvement, are recommended.

Crash severity analysis of vulnerable road users using machine learning

Road crash fatality is a universal problem of the transportation system. A massive death toll caused annually due to road crash incidents, and among them, vulnerable road users (VRU) are endangered with high crash severity. This paper focuses on employing machine learning-based classification approaches for modelling injury severity of vulnerable road users—pedestrian, bicyclist, and motorcyclist. Specifically, this study aims to analyse critical features associated with different VRU groups—for pedestrian, bicyclist, motorcyclist and all VRU groups together. The critical factor of crash severity outcomes for these VRU groups is estimated in identifying the similarities and differences across different important features associated with different VRU groups. The crash data for the study is sourced from the state of Queensland in Australia for the years 2013 through 2019. The supervised machine learning algorithms considered for the empirical analysis includes the K-Nearest Neighbour (KNN), Support Vector Machine (SVM) and Random Forest (RF). In these models, 17 distinct road crash parameters are considered as input features to train models, which originate from road user characteristics, weather and environment, vehicle and driver condition, period, road characteristics and regions, traffic, and speed jurisdiction. These classification models are separately trained and tested for individual and unified VRU to assess crash severity levels. Afterwards, model performances are compared with each other to justify the best classifier where Random Forest classification models for all VRU modes are found to be comparatively robust in test accuracy: (motorcyclist: 72.30%, bicyclist: 64.45%, pedestrian: 67.23%, unified VRU: 68.57%). Based on the Random Forest model, the road crash features are ranked and compared according to their impact on crash severity classification. Furthermore, a model-based partial dependency of each road crash parameters on the severity levels is plotted and compared for each individual and unified VRU. This clarifies the tendency of road crash parameters to vary with different VRU crash severity. Based on the outcome of the comparative analysis, motorcyclists are found to be more likely exposed to higher crash severity, followed by pedestrians and bicyclists.