Search Results (14)

This paper provides a systematic review of recent applications of NLP methods for analyzing traffic crash reports, with a focus on estimating crash severity, crash duration, and crash causation. The review covers prior research using probabilistic topic modeling methods such as LDA, STM, and hierarchical Dirichlet processes in addition to research using transformer-based language models, which include encoder-based models like BERT and PubMedBERT as well as decoder-based models like GPT, GPT2, ChatGPT, GPT-3, and LLaMA. The review starts with a systematic literature selection process with predefined inclusion criteria. We categorize the reviewed studies into the following application areas: crash severity prediction, risk factor identification in crashes, and road safety analysis. The results show several complementary advantages of using different NLP techniques to achieve different analytical goals. Topic models allow for interpretable and exploratory pattern discovery, while encoder models are well-suited for structured prediction problems. Decoder models have the additional flexibility to perform zero-shot and few-shot reasoning, which makes them useful for reasoning about under-sampled or under-reported data. Across the literature, hybrid methods that combine text and structured data outperform individual methods in terms of prediction accuracy and broad applicability. Challenges across the literature include class imbalance, lack of standardization in preprocessing and evaluation methods, and the tradeoff between prediction accuracy and interpretability of prediction models. These findings highlight the importance of aligning model selection with data availability and operational constraints, pointing toward future research directions in hybrid modeling frameworks, standardized evaluation protocols, and real-world deployment of NLP-driven traffic safety systems. Full article

Miniature commercial trucks constitute a critical component of urban freight systems but face elevated crash risk due to distinctive driving patterns, frequent operation, and variable loads. This study quantifies how long-term and short-term driving behaviors jointly shape crash economic loss levels and identifies factors most strongly associated with severe claims. A driver-level dataset linking multi-source running behavior indicators, vehicle attributes, and insurance claims is constructed, and an enhanced Wasserstein generative adversarial network with Euclidean distance is employed to synthesize minority crash samples and alleviate class imbalance. Crash economic loss levels are modeled using a random-effects generalized ordinal logit specification, and model performance is compared with a generalized ordered logit benchmark. Marginal effects analysis is used to evaluate the influence of pre-collision driving states (straight, turning, reversing, rolling, following closely) and key behavioral indicators. Results indicate significant effects of inter-provincial duration and count ratios, morning and empty-trip frequencies, no-claim discount coefficients, and vehicle age on crash economic loss, with prolonged speeding duration and fatigued mileage associated with major losses, whereas frequent speeding and fatigue episodes are primarily linked to minor claims. These findings clarify causal patterns for miniature commercial truck crashes with different economic losses and provide an empirical basis for targeted safety interventions and refined insurance pricing. Full article

This study investigates the causes of traffic accidents involving Advanced Driver Assistance Systems (ADAS) and Autonomous Driving Systems (ADS) and their interdependencies. Using a source dataset comprising 3015 ADAS accident records and 1085 ADS accident records from National Highway Traffic Safety Administration (NHTSA), the study categorizes accident severity into four levels and applies association rule mining (ARM) to identify high-frequency risk factor combinations. Key risk factors include environmental, road, vehicle, and accident characteristics. Findings show that ADAS accidents are concentrated in highway straight-driving scenarios, strongly correlated with rainy weather, and often involve rear-end collisions due to delayed driver reactions. ADS accidents predominantly occur in intersection stopping scenarios, favor clear weather, and exhibit better safety performance in non-damage cases with Level 5 (L5) systems, though they still face perception and decision-making challenges in complex scenarios like nighttime wet roads. The study further reveals that vehicle design purpose (ADAS for highways, L5 for urban areas) strongly influences accident severity, with L5 systems reducing fatality risks through advanced perception but still affected by high speeds, extreme lighting, and system aging. Make attributes and technological maturity also significantly impact outcomes. This study provides insights for technological advancement, regulatory improvements, and human–machine collaboration optimization. Full article

Predicting road crash severity remains a major challenge in transportation safety research, requiring models that combine predictive accuracy, interpretability, and computational efficiency. This study introduces a Multi-Risk Index based on Copula Integration (MRI-Copula)—a hybrid framework that integrates Categorical Boosting (CatBoost) with SHapley Additive exPlanations (SHAP) and Vine Copula dependence modeling to assess and predict crash severity. The approach leverages CatBoost–SHAP to quantify the marginal contribution of each risk factor while maintaining model transparency and employs copula-based tail dependence to capture the joint escalation of risk under extreme crash conditions. Using a dataset of 877 police-reported crashes from Jeddah, Saudi Arabia, the framework constructs three interpretable sub-indices—Environmental Risk Index (ERI), Behavioural Risk Index (BRI), and Systemic Risk Index (SRI)—representing distinct domains of crash causation. These indices are combined through a convex weighting parameter (α), optimized via cross-validation (optimal α = 0.80), ensuring a balanced integration of predictive and dependence-based information. Comparative evaluation across multiple classifiers—CatBoost, Light Gradient Boosting Machine (LightGBM), Histogram-based Gradient Boosting (HistGB), and Logistic Regression—demonstrated the robustness of the framework. The CatBoost + MRI-Copula configuration achieved the highest predictive performance (AUC = 0.986; F1 = 0.904), while LightGBM and HistGB offered comparable accuracy (AUC ≈ 0.958; F1 ≈ 0.89) at a fraction of the computational time (≤1 s versus 32 s for CatBoost), highlighting a trade-off between analytical precision and scalability. Consequently, the MRI-Copula framework provides a transparent and theoretically grounded foundation for data-driven road safety management. It bridges predictive analytics and decision support offering a scalable, interpretable, and policy-relevant tool for proactive crash risk mitigation. Full article

Road traffic crashes pose significant economic and public health burdens, necessitating an in-depth understanding of crash causation and its links to underlying factors. This study introduces a machine learning-based hurdle model framework tailored for analyzing zero-inflated crash frequency data, addressing the limitations of traditional statistical models like the Poisson and negative binomial models, which struggle with zero-inflation and overdispersion. The research employs a two-stage modeling process using CatBoost. The first stage uses binary classification to identify road segments with potential crash occurrences, applying a customized loss function to tackle data imbalance. The second stage predicts crash frequency, also utilizing a customized loss function for count data. SHapley Additive exPlanations (SHAP) analysis interprets the model outcomes, providing insights into factors affecting crash likelihood and frequency. This study validates the model’s performance with real-world crash data from 2011 to 2015 in South Korea, demonstrating superior accuracy in both the classification and regression stages compared to other machine learning algorithms and traditional models. These findings have significant implications for traffic safety research and policymaking, offering stakeholders a more accurate and interpretable tool for crash data analysis to develop targeted safety interventions. Full article

Governments and authorities worldwide consider road traffic crashes (RTCs) to be a major concern. These crashes incur losses in terms of productivity, property, and life. For a country to establish its road and action plans, it is crucial to comprehend the reasons for and consequences of traffic collisions. The main objective of this research study was to evaluate and rank the important and supporting factors influencing traffic crashes on the road. To identify the most significant accident causation elements, the proportion-based analytic hierarchy process (PBAHP) was used to order the factors in terms of their relative importance. In this study, the city of Al-Ahsa, located in the eastern province of Saudi Arabia, was used as a case study, since this city is the highest RTC-prone area in the region. PBAHP was used to calculate relative importance/weights for different crash types and reasons in terms of their impact on crash severity. It was found that vehicle-overturned collisions which result in fatal crashes have the most weight, whereas “hit motorcycle” crashes result in serious injury crashes. When vehicles (two or more) collide with one another while they are moving, it appears that the likelihood of a fatality in a collision increases. The highest weights for serious injury crashes came from “driver distraction”, “leaving insufficient safe distance”, and “speeding”, which also generated similar and relatively high weights for fatal crashes. Weights from the PBAHP approach were also used to develop utility functions for predicting the severity of crashes. This approach could assist decision-makers in concentrating on the key elements affecting road traffic crashes and enhancing road safety. Full article

Traffic crashes involving pedestrians have a high frequency in developing countries. Among road users, pedestrians are the most vulnerable, as their involvement in traffic crashes is usually followed by severe and fatal injuries. This study aims to identify pedestrian crash patterns and reveal the random parameters in the dataset. A three-year (2015–2017) pedestrian crash dataset in Mashhad, Iran, was employed to investigate the influence of a rich set of factors on pedestrian injury severity, some of which have been less accounted for in previous studies (e.g., the vicinity to overpasses, the existence of vegetated buffers, and park lanes). A two-step method integrating latent class cluster analysis (LCA) and the mixed logit model was utilized to consider unobserved heterogeneity. The results demonstrated that various factors related to the pedestrian, vehicle, temporal, environmental, roadway, and built-environment characteristics are associated with pedestrian injuries. Furthermore, it was found that integrated use of LCA and mixed logit models can considerably reduce the unobserved heterogeneity and uncover the hidden effects influencing severity outcomes, leading to a more profound perception of pedestrian crash causation. The findings of this research can act as a helpful resource for implementing effective strategies by policymakers to reduce pedestrian casualties. Full article

Heavy goods vehicles (HGVs) are involved in 4.5% of police-reported road crashes in Europe and 14.2% of fatal road crashes. Active and passive safety systems can help to prevent crashes or mitigate the consequences but need detailed scenarios based on analysis of region-specific data to be designed effectively; however, a sufficiently detailed overview focusing on long-haul trucks is not available for Europe. The aim of this paper is to give a comprehensive and up-to-date analysis of crashes in the European Union that involve HGVs weighing 16 tons or more (16 t+). The identification of the most critical scenarios and their characteristics is based on a three-level analysis, as follows. Crash statistics based on data from the Community Database on Accidents on the Roads in Europe (CARE) provide a general overview of crashes involving HGVs. These results are complemented by a more detailed characterization of crashes involving 16 t+ trucks based on national road crash data from Italy, Spain, and Sweden. This analysis is further refined by a detailed study of crashes involving 16 t+ trucks in the German In-Depth Accident Study (GIDAS), including a crash causation analysis. The results show that most European HGV crashes occur in clear weather, during daylight, on dry roads, outside city limits, and on nonhighway roads. Three main scenarios for 16 t+ trucks are characterized in-depth: rear-end crashes in which the truck is the striking partner, conflicts during right turn maneuvers of the truck with a cyclist riding alongside, and pedestrians crossing the road in front of the truck. Among truck-related crash causes, information admission failures (e.g., distraction) were the main crash causation factor in 72% of cases in the rear-end striking scenario while information access problems (e.g., blind spots) were present for 72% of cases in the cyclist scenario and 75% of cases in the pedestrian scenario. The three levels of data analysis used in this paper give a deeper understanding of European HGV crashes, in terms of the most common crash characteristics on EU level and very detailed descriptions of both kinematic parameters and crash causation factors for the above scenarios. The results thereby provide both a global overview and sufficient depth of analysis of the most relevant cases and aid safety system development. Full article

Traffic crashes are a common cause of injury and death, and often result from the negligent actions of an inattentive, speeding, or impaired driver. In such cases, a civil legal action may be brought by an injured claimant for compensation for injuries resulting from a crash. Crash-related litigation is defended on various theories, one of which is to raise the issue of contributory negligence when the claimant was not using an available seat belt at the time of the crash, based on the assertion that the claimed injuries would have been avoided or minimized to some degree if the claimant had been restrained. At present, there are no published standards or systematic approach for assessing and quantifying the contribution of seat belt non-use to the cause of a claimant’s specific injury. A reliable medicolegal analysis that addresses whether contributory negligence can be proven in a specific case requires a multidisciplinary approach: First, the nature and severity of the crash must be reconstructed as it affected the vehicle kinetics (engineering) and in turn affected the kinematics of the occupant (biomechanics), next, the injuries must be described and scaled for severity (medicine/pathology), and finally, the risk of the known injuries given the actual circumstances of the crash and occupant (i.e., unbelted) are compared to the risk of the same injuries, and the same crash circumstances, but in the hypothetical scenario in which the claimant is belted. In the present discussion, methods for analyzing the presence and quantifying the degree of contributory negligence for seat belt non-use, suitable for presentation in a medicolegal setting, are described and illustrated with an example from the author’s personal case inventory. A detailed reconstruction of the crash is described, along with the associated occupant kinematics, and the resulting observed injuries. The injuries are then categorized by their anatomical location, type, and severity using Abbreviated Injury Scale designations. Quantification of the injury risk for the actual (unbelted) vs. hypothetical (belted) scenario is based on case-specific analysis of data accessed from a US national crash injury database The difference in risk for the two exposure scenarios can be quantified in terms of either relative risk (a risk ratio) or attributable risk (a risk proportion), with the goal to determine whether the analysis meets the threshold of a relative risk of >2.0, or an attributable risk of 50%, in order to meet the “more probable than not” standard typically required by courts. As a final step in a reliable analysis that exceeds the legal threshold for relevant evidence, the absolute increase in risk is used to quantify the degree to which the claimant’s seat belt non-use contributed to the likelihood of their injuries. Full article

Injury claims associated with minimal damage rear impact traffic crashes are often defended using a “biomechanical approach,” in which the occupant forces of the crash are compared to the forces of activities of daily living (ADLs), resulting in the conclusion that the risk of injury from the crash is the same as for ADLs. The purpose of the present investigation is to evaluate the scientific validity of the central operating premise of the biomechanical approach to injury causation; that occupant acceleration is a scientifically valid proxy for injury risk. Data were abstracted, pooled, and compared from three categories of published literature: (1) volunteer rear impact crash testing studies, (2) ADL studies, and (3) observational studies of real-world rear impacts. We compared the occupant accelerations of minimal or no damage (i.e., 3 to 11 kph speed change or “delta V”) rear impact crash tests to the accelerations described in 6 of the most commonly reported ADLs in the reviewed studies. As a final step, the injury risk observed in real world crashes was compared to the results of the pooled crash test and ADL analyses, controlling for delta V. The results of the analyses indicated that average peak linear and angular acceleration forces observed at the head during rear impact crash tests were typically at least several times greater than average forces observed during ADLs. In contrast, the injury risk of real-world minimal damage rear impact crashes was estimated to be at least 2000 times greater than for any ADL. The results of our analysis indicate that the principle underlying the biomechanical injury causation approach, that occupant acceleration is a proxy for injury risk, is scientifically invalid. The biomechanical approach to injury causation in minimal damage crashes invariably results in the vast underestimation of the actual risk of such crashes, and should be discontinued as it is a scientifically invalid practice. Full article

Road traffic crashes cause fatalities and injuries of both drivers/passengers in vehicles and pedestrians outside, thus challenge public health especially in big cities in developing countries like China. Previous efforts mainly focus on a specific crash type or causation to examine the crash characteristics in China while lacking the characteristics of various crash types, factors, and the interplay between them. This study investigated the crash characteristics in Shenzhen, one of the biggest four cities in China, based on the police-reported crashes from 2014 to 2016. The descriptive characteristics were reported in detail with respect to each of the crash attributes. Based on the recorded crash locations, the land-use pattern was obtained as one of the attributes for each crash. Then, the relationship between the attributes in motor-vehicle-involved crashes was examined using the Bayesian network analysis. We revealed the distinct crash characteristics observed between the examined levels of each attribute, as well the interplay between the attributes. This study provides an insight into the crash characteristics in Shenzhen, which would help understand the driving behavior of Chinese drivers, identify the traffic safety problems, guide the research focuses on advanced driver assistance systems (ADASs) and traffic management countermeasures in China. Full article

Understanding the association between crash attributes and drivers’ crash involvement in different types of crashes can help figure out the causation of crashes. The aim of this study was to examine the involvement in different types of crashes for drivers from different age groups, by using the police-reported crash data from 2014 to 2016 in Shenzhen, China. A synthetic minority oversampling technique (SMOTE) together with edited nearest neighbors (ENN) were used to solve the data imbalance problem caused by the lack of crash records of older drivers. Logistic regression was utilized to estimate the probability of a certain type of crashes, and odds ratios that were calculated based on the logistic regression results were used to quantify the association between crash attributes and drivers’ crash involvement in different types of crashes. Results showed that drivers’ involvement patterns in different crash types were affected by different factors, and the involvement patterns differed among the examined age groups. Knowledge generated from the present study could help improve the development of countermeasures for driving safety enhancement. Full article

Road trauma is a significant health problem in rural and remote regions of Australia, particularly for Indigenous communities. This study aims to identify and compare the circumstances leading to (proximal causation) and social determinants of (distal causation) crashes of Indigenous and non-Indigenous people in these regions and their relation to remoteness. This is a topic seriously under-researched in Australia. Modelled on an earlier study, 229 persons injured in crashes were recruited from local health facilities in rural and remote North Queensland and interviewed, mainly by telephone, according to a fixed protocol which included a detailed narrative of the circumstances of the crash. A qualitative analysis of these narratives identified several core themes, further explored statistically in this sample, supplemented by participants in the earlier study with compatible questionnaire data, designed to determine which factors were more closely associated with Indigenous status and which with remoteness. Indigenous participants were less often vehicle controllers, more likely to have recently been a drink driver or passenger thereof; to be unemployed, unlicensed, distracted or fatigued before the crash, alcohol dependent and have lower perceived social, but not personal, locus of control in a traffic crash than non-Indigenous persons. Differences between Indigenous and non-Indigenous participants are largely ascribable to hardship and transport disadvantage due to lack of access to licensing and associated limitations on employment opportunities. Based on these findings, a number of policy recommendations relating to educational, enforcement and engineering issues have been made. Full article

Road traffic safety is the result of a complex interaction of factors, and causes behind road vehicle crashes require different measures to reduce their impacts. This study assesses how strongly the variation in daily winter crash rates associates with weather conditions in Finland. This is done by illustrating trends and spatiotemporal variation in the crash rates, by showing how a GIS application can evidence the association between temporary rises in regional crash rates and the occurrence of bad weather, and with a regression model on crash rate sensitivity to adverse weather conditions. The analysis indicates that a base rate of crashes depending on non-weather factors exists, and some combinations of extreme weather conditions are able to substantially push up crash rates on days with bad weather. Some spatial causation factors, such as variation of geophysical characteristics causing systematic differences in the distributions of weather variables, exist. Yet, even in winter, non-spatial factors are normally more significant. GIS data can support optimal deployment of rescue services and enhance in-depth quantitative analysis by helping to identify the most appropriate spatial and temporal resolutions. However, the supportive role of GIS should not be inferred as existence of highly significant spatial causation. Full article